1 {-| Implementation of cluster-wide logic.
3 This module holds all pure cluster-logic; I\/O related functionality
4 goes into the /Main/ module for the individual binaries.
10 Copyright (C) 2009, 2010, 2011, 2012 Google Inc.
12 This program is free software; you can redistribute it and/or modify
13 it under the terms of the GNU General Public License as published by
14 the Free Software Foundation; either version 2 of the License, or
15 (at your option) any later version.
17 This program is distributed in the hope that it will be useful, but
18 WITHOUT ANY WARRANTY; without even the implied warranty of
19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 General Public License for more details.
22 You should have received a copy of the GNU General Public License
23 along with this program; if not, write to the Free Software
24 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
29 module Ganeti.HTools.Cluster
38 -- * Generic functions
40 , computeAllocationDelta
41 -- * First phase functions
43 -- * Second phase functions
48 -- * Display functions
51 -- * Balacing functions
60 -- * IAllocator functions
67 -- * Allocation functions
70 -- * Node group functions
76 import qualified Data.IntSet as IntSet
78 import Data.Maybe (fromJust, isNothing)
79 import Data.Ord (comparing)
80 import Text.Printf (printf)
82 import qualified Ganeti.HTools.Container as Container
83 import qualified Ganeti.HTools.Instance as Instance
84 import qualified Ganeti.HTools.Node as Node
85 import qualified Ganeti.HTools.Group as Group
86 import Ganeti.HTools.Types
87 import Ganeti.HTools.Utils
88 import Ganeti.HTools.Compat
89 import qualified Ganeti.OpCodes as OpCodes
93 -- | Allocation\/relocation solution.
94 data AllocSolution = AllocSolution
95 { asFailures :: [FailMode] -- ^ Failure counts
96 , asAllocs :: Int -- ^ Good allocation count
97 , asSolution :: Maybe Node.AllocElement -- ^ The actual allocation result
98 , asLog :: [String] -- ^ Informational messages
101 -- | Node evacuation/group change iallocator result type. This result
102 -- type consists of actual opcodes (a restricted subset) that are
103 -- transmitted back to Ganeti.
104 data EvacSolution = EvacSolution
105 { esMoved :: [(Idx, Gdx, [Ndx])] -- ^ Instances moved successfully
106 , esFailed :: [(Idx, String)] -- ^ Instances which were not
108 , esOpCodes :: [[OpCodes.OpCode]] -- ^ List of jobs
111 -- | Allocation results, as used in 'iterateAlloc' and 'tieredAlloc'.
112 type AllocResult = (FailStats, Node.List, Instance.List,
113 [Instance.Instance], [CStats])
115 -- | A type denoting the valid allocation mode/pairs.
117 -- For a one-node allocation, this will be a @Left ['Ndx']@, whereas
118 -- for a two-node allocation, this will be a @Right [('Ndx',
119 -- ['Ndx'])]@. In the latter case, the list is basically an
120 -- association list, grouped by primary node and holding the potential
121 -- secondary nodes in the sub-list.
122 type AllocNodes = Either [Ndx] [(Ndx, [Ndx])]
124 -- | The empty solution we start with when computing allocations.
125 emptyAllocSolution :: AllocSolution
126 emptyAllocSolution = AllocSolution { asFailures = [], asAllocs = 0
127 , asSolution = Nothing, asLog = [] }
129 -- | The empty evac solution.
130 emptyEvacSolution :: EvacSolution
131 emptyEvacSolution = EvacSolution { esMoved = []
136 -- | The complete state for the balancing solution.
137 data Table = Table Node.List Instance.List Score [Placement]
138 deriving (Show, Read)
140 -- | Cluster statistics data type.
142 { csFmem :: Integer -- ^ Cluster free mem
143 , csFdsk :: Integer -- ^ Cluster free disk
144 , csAmem :: Integer -- ^ Cluster allocatable mem
145 , csAdsk :: Integer -- ^ Cluster allocatable disk
146 , csAcpu :: Integer -- ^ Cluster allocatable cpus
147 , csMmem :: Integer -- ^ Max node allocatable mem
148 , csMdsk :: Integer -- ^ Max node allocatable disk
149 , csMcpu :: Integer -- ^ Max node allocatable cpu
150 , csImem :: Integer -- ^ Instance used mem
151 , csIdsk :: Integer -- ^ Instance used disk
152 , csIcpu :: Integer -- ^ Instance used cpu
153 , csTmem :: Double -- ^ Cluster total mem
154 , csTdsk :: Double -- ^ Cluster total disk
155 , csTcpu :: Double -- ^ Cluster total cpus
156 , csVcpu :: Integer -- ^ Cluster total virtual cpus
157 , csNcpu :: Double -- ^ Equivalent to 'csIcpu' but in terms of
158 -- physical CPUs, i.e. normalised used phys CPUs
159 , csXmem :: Integer -- ^ Unnacounted for mem
160 , csNmem :: Integer -- ^ Node own memory
161 , csScore :: Score -- ^ The cluster score
162 , csNinst :: Int -- ^ The total number of instances
163 } deriving (Show, Read)
165 -- | A simple type for allocation functions.
166 type AllocMethod = Node.List -- ^ Node list
167 -> Instance.List -- ^ Instance list
168 -> Maybe Int -- ^ Optional allocation limit
169 -> Instance.Instance -- ^ Instance spec for allocation
170 -> AllocNodes -- ^ Which nodes we should allocate on
171 -> [Instance.Instance] -- ^ Allocated instances
172 -> [CStats] -- ^ Running cluster stats
173 -> Result AllocResult -- ^ Allocation result
175 -- * Utility functions
177 -- | Verifies the N+1 status and return the affected nodes.
178 verifyN1 :: [Node.Node] -> [Node.Node]
179 verifyN1 = filter Node.failN1
181 {-| Computes the pair of bad nodes and instances.
183 The bad node list is computed via a simple 'verifyN1' check, and the
184 bad instance list is the list of primary and secondary instances of
188 computeBadItems :: Node.List -> Instance.List ->
189 ([Node.Node], [Instance.Instance])
190 computeBadItems nl il =
191 let bad_nodes = verifyN1 $ getOnline nl
192 bad_instances = map (`Container.find` il) .
194 concatMap (\ n -> Node.sList n ++ Node.pList n) bad_nodes
196 (bad_nodes, bad_instances)
198 -- | Extracts the node pairs for an instance. This can fail if the
199 -- instance is single-homed. FIXME: this needs to be improved,
200 -- together with the general enhancement for handling non-DRBD moves.
201 instanceNodes :: Node.List -> Instance.Instance ->
202 (Ndx, Ndx, Node.Node, Node.Node)
203 instanceNodes nl inst =
204 let old_pdx = Instance.pNode inst
205 old_sdx = Instance.sNode inst
206 old_p = Container.find old_pdx nl
207 old_s = Container.find old_sdx nl
208 in (old_pdx, old_sdx, old_p, old_s)
210 -- | Zero-initializer for the CStats type.
211 emptyCStats :: CStats
212 emptyCStats = CStats 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
214 -- | Update stats with data from a new node.
215 updateCStats :: CStats -> Node.Node -> CStats
216 updateCStats cs node =
217 let CStats { csFmem = x_fmem, csFdsk = x_fdsk,
218 csAmem = x_amem, csAcpu = x_acpu, csAdsk = x_adsk,
219 csMmem = x_mmem, csMdsk = x_mdsk, csMcpu = x_mcpu,
220 csImem = x_imem, csIdsk = x_idsk, csIcpu = x_icpu,
221 csTmem = x_tmem, csTdsk = x_tdsk, csTcpu = x_tcpu,
222 csVcpu = x_vcpu, csNcpu = x_ncpu,
223 csXmem = x_xmem, csNmem = x_nmem, csNinst = x_ninst
226 inc_amem = Node.fMem node - Node.rMem node
227 inc_amem' = if inc_amem > 0 then inc_amem else 0
228 inc_adsk = Node.availDisk node
229 inc_imem = truncate (Node.tMem node) - Node.nMem node
230 - Node.xMem node - Node.fMem node
231 inc_icpu = Node.uCpu node
232 inc_idsk = truncate (Node.tDsk node) - Node.fDsk node
233 inc_vcpu = Node.hiCpu node
234 inc_acpu = Node.availCpu node
235 inc_ncpu = fromIntegral (Node.uCpu node) /
236 iPolicyVcpuRatio (Node.iPolicy node)
237 in cs { csFmem = x_fmem + fromIntegral (Node.fMem node)
238 , csFdsk = x_fdsk + fromIntegral (Node.fDsk node)
239 , csAmem = x_amem + fromIntegral inc_amem'
240 , csAdsk = x_adsk + fromIntegral inc_adsk
241 , csAcpu = x_acpu + fromIntegral inc_acpu
242 , csMmem = max x_mmem (fromIntegral inc_amem')
243 , csMdsk = max x_mdsk (fromIntegral inc_adsk)
244 , csMcpu = max x_mcpu (fromIntegral inc_acpu)
245 , csImem = x_imem + fromIntegral inc_imem
246 , csIdsk = x_idsk + fromIntegral inc_idsk
247 , csIcpu = x_icpu + fromIntegral inc_icpu
248 , csTmem = x_tmem + Node.tMem node
249 , csTdsk = x_tdsk + Node.tDsk node
250 , csTcpu = x_tcpu + Node.tCpu node
251 , csVcpu = x_vcpu + fromIntegral inc_vcpu
252 , csNcpu = x_ncpu + inc_ncpu
253 , csXmem = x_xmem + fromIntegral (Node.xMem node)
254 , csNmem = x_nmem + fromIntegral (Node.nMem node)
255 , csNinst = x_ninst + length (Node.pList node)
258 -- | Compute the total free disk and memory in the cluster.
259 totalResources :: Node.List -> CStats
261 let cs = foldl' updateCStats emptyCStats . Container.elems $ nl
262 in cs { csScore = compCV nl }
264 -- | Compute the delta between two cluster state.
266 -- This is used when doing allocations, to understand better the
267 -- available cluster resources. The return value is a triple of the
268 -- current used values, the delta that was still allocated, and what
269 -- was left unallocated.
270 computeAllocationDelta :: CStats -> CStats -> AllocStats
271 computeAllocationDelta cini cfin =
272 let CStats {csImem = i_imem, csIdsk = i_idsk, csIcpu = i_icpu,
273 csNcpu = i_ncpu } = cini
274 CStats {csImem = f_imem, csIdsk = f_idsk, csIcpu = f_icpu,
275 csTmem = t_mem, csTdsk = t_dsk, csVcpu = f_vcpu,
276 csNcpu = f_ncpu, csTcpu = f_tcpu } = cfin
277 rini = AllocInfo { allocInfoVCpus = fromIntegral i_icpu
278 , allocInfoNCpus = i_ncpu
279 , allocInfoMem = fromIntegral i_imem
280 , allocInfoDisk = fromIntegral i_idsk
282 rfin = AllocInfo { allocInfoVCpus = fromIntegral (f_icpu - i_icpu)
283 , allocInfoNCpus = f_ncpu - i_ncpu
284 , allocInfoMem = fromIntegral (f_imem - i_imem)
285 , allocInfoDisk = fromIntegral (f_idsk - i_idsk)
287 runa = AllocInfo { allocInfoVCpus = fromIntegral (f_vcpu - f_icpu)
288 , allocInfoNCpus = f_tcpu - f_ncpu
289 , allocInfoMem = truncate t_mem - fromIntegral f_imem
290 , allocInfoDisk = truncate t_dsk - fromIntegral f_idsk
292 in (rini, rfin, runa)
294 -- | The names and weights of the individual elements in the CV list.
295 detailedCVInfo :: [(Double, String)]
296 detailedCVInfo = [ (1, "free_mem_cv")
297 , (1, "free_disk_cv")
299 , (1, "reserved_mem_cv")
300 , (4, "offline_all_cnt")
301 , (16, "offline_pri_cnt")
302 , (1, "vcpu_ratio_cv")
305 , (1, "disk_load_cv")
307 , (2, "pri_tags_score")
310 -- | Holds the weights used by 'compCVNodes' for each metric.
311 detailedCVWeights :: [Double]
312 detailedCVWeights = map fst detailedCVInfo
314 -- | Compute the mem and disk covariance.
315 compDetailedCV :: [Node.Node] -> [Double]
316 compDetailedCV all_nodes =
317 let (offline, nodes) = partition Node.offline all_nodes
318 mem_l = map Node.pMem nodes
319 dsk_l = map Node.pDsk nodes
320 -- metric: memory covariance
321 mem_cv = stdDev mem_l
322 -- metric: disk covariance
323 dsk_cv = stdDev dsk_l
324 -- metric: count of instances living on N1 failing nodes
325 n1_score = fromIntegral . sum . map (\n -> length (Node.sList n) +
326 length (Node.pList n)) .
327 filter Node.failN1 $ nodes :: Double
328 res_l = map Node.pRem nodes
329 -- metric: reserved memory covariance
330 res_cv = stdDev res_l
331 -- offline instances metrics
332 offline_ipri = sum . map (length . Node.pList) $ offline
333 offline_isec = sum . map (length . Node.sList) $ offline
334 -- metric: count of instances on offline nodes
335 off_score = fromIntegral (offline_ipri + offline_isec)::Double
336 -- metric: count of primary instances on offline nodes (this
337 -- helps with evacuation/failover of primary instances on
338 -- 2-node clusters with one node offline)
339 off_pri_score = fromIntegral offline_ipri::Double
340 cpu_l = map Node.pCpu nodes
341 -- metric: covariance of vcpu/pcpu ratio
342 cpu_cv = stdDev cpu_l
343 -- metrics: covariance of cpu, memory, disk and network load
344 (c_load, m_load, d_load, n_load) =
346 let DynUtil c1 m1 d1 n1 = Node.utilLoad n
347 DynUtil c2 m2 d2 n2 = Node.utilPool n
348 in (c1/c2, m1/m2, d1/d2, n1/n2)) nodes
349 -- metric: conflicting instance count
350 pri_tags_inst = sum $ map Node.conflictingPrimaries nodes
351 pri_tags_score = fromIntegral pri_tags_inst::Double
352 in [ mem_cv, dsk_cv, n1_score, res_cv, off_score, off_pri_score, cpu_cv
353 , stdDev c_load, stdDev m_load , stdDev d_load, stdDev n_load
356 -- | Compute the /total/ variance.
357 compCVNodes :: [Node.Node] -> Double
358 compCVNodes = sum . zipWith (*) detailedCVWeights . compDetailedCV
360 -- | Wrapper over 'compCVNodes' for callers that have a 'Node.List'.
361 compCV :: Node.List -> Double
362 compCV = compCVNodes . Container.elems
364 -- | Compute online nodes from a 'Node.List'.
365 getOnline :: Node.List -> [Node.Node]
366 getOnline = filter (not . Node.offline) . Container.elems
368 -- * Balancing functions
370 -- | Compute best table. Note that the ordering of the arguments is important.
371 compareTables :: Table -> Table -> Table
372 compareTables a@(Table _ _ a_cv _) b@(Table _ _ b_cv _ ) =
373 if a_cv > b_cv then b else a
375 -- | Applies an instance move to a given node list and instance.
376 applyMove :: Node.List -> Instance.Instance
377 -> IMove -> OpResult (Node.List, Instance.Instance, Ndx, Ndx)
379 applyMove nl inst Failover =
380 let (old_pdx, old_sdx, old_p, old_s) = instanceNodes nl inst
381 int_p = Node.removePri old_p inst
382 int_s = Node.removeSec old_s inst
383 new_nl = do -- Maybe monad
384 new_p <- Node.addPriEx (Node.offline old_p) int_s inst
385 new_s <- Node.addSec int_p inst old_sdx
386 let new_inst = Instance.setBoth inst old_sdx old_pdx
387 return (Container.addTwo old_pdx new_s old_sdx new_p nl,
388 new_inst, old_sdx, old_pdx)
391 -- Replace the primary (f:, r:np, f)
392 applyMove nl inst (ReplacePrimary new_pdx) =
393 let (old_pdx, old_sdx, old_p, old_s) = instanceNodes nl inst
394 tgt_n = Container.find new_pdx nl
395 int_p = Node.removePri old_p inst
396 int_s = Node.removeSec old_s inst
397 force_p = Node.offline old_p
398 new_nl = do -- Maybe monad
399 -- check that the current secondary can host the instance
400 -- during the migration
401 tmp_s <- Node.addPriEx force_p int_s inst
402 let tmp_s' = Node.removePri tmp_s inst
403 new_p <- Node.addPriEx force_p tgt_n inst
404 new_s <- Node.addSecEx force_p tmp_s' inst new_pdx
405 let new_inst = Instance.setPri inst new_pdx
406 return (Container.add new_pdx new_p $
407 Container.addTwo old_pdx int_p old_sdx new_s nl,
408 new_inst, new_pdx, old_sdx)
411 -- Replace the secondary (r:ns)
412 applyMove nl inst (ReplaceSecondary new_sdx) =
413 let old_pdx = Instance.pNode inst
414 old_sdx = Instance.sNode inst
415 old_s = Container.find old_sdx nl
416 tgt_n = Container.find new_sdx nl
417 int_s = Node.removeSec old_s inst
418 force_s = Node.offline old_s
419 new_inst = Instance.setSec inst new_sdx
420 new_nl = Node.addSecEx force_s tgt_n inst old_pdx >>=
421 \new_s -> return (Container.addTwo new_sdx
422 new_s old_sdx int_s nl,
423 new_inst, old_pdx, new_sdx)
426 -- Replace the secondary and failover (r:np, f)
427 applyMove nl inst (ReplaceAndFailover new_pdx) =
428 let (old_pdx, old_sdx, old_p, old_s) = instanceNodes nl inst
429 tgt_n = Container.find new_pdx nl
430 int_p = Node.removePri old_p inst
431 int_s = Node.removeSec old_s inst
432 force_s = Node.offline old_s
433 new_nl = do -- Maybe monad
434 new_p <- Node.addPri tgt_n inst
435 new_s <- Node.addSecEx force_s int_p inst new_pdx
436 let new_inst = Instance.setBoth inst new_pdx old_pdx
437 return (Container.add new_pdx new_p $
438 Container.addTwo old_pdx new_s old_sdx int_s nl,
439 new_inst, new_pdx, old_pdx)
442 -- Failver and replace the secondary (f, r:ns)
443 applyMove nl inst (FailoverAndReplace new_sdx) =
444 let (old_pdx, old_sdx, old_p, old_s) = instanceNodes nl inst
445 tgt_n = Container.find new_sdx nl
446 int_p = Node.removePri old_p inst
447 int_s = Node.removeSec old_s inst
448 force_p = Node.offline old_p
449 new_nl = do -- Maybe monad
450 new_p <- Node.addPriEx force_p int_s inst
451 new_s <- Node.addSecEx force_p tgt_n inst old_sdx
452 let new_inst = Instance.setBoth inst old_sdx new_sdx
453 return (Container.add new_sdx new_s $
454 Container.addTwo old_sdx new_p old_pdx int_p nl,
455 new_inst, old_sdx, new_sdx)
458 -- | Tries to allocate an instance on one given node.
459 allocateOnSingle :: Node.List -> Instance.Instance -> Ndx
460 -> OpResult Node.AllocElement
461 allocateOnSingle nl inst new_pdx =
462 let p = Container.find new_pdx nl
463 new_inst = Instance.setBoth inst new_pdx Node.noSecondary
465 Instance.instMatchesPolicy inst (Node.iPolicy p)
466 new_p <- Node.addPri p inst
467 let new_nl = Container.add new_pdx new_p nl
468 new_score = compCV nl
469 return (new_nl, new_inst, [new_p], new_score)
471 -- | Tries to allocate an instance on a given pair of nodes.
472 allocateOnPair :: Node.List -> Instance.Instance -> Ndx -> Ndx
473 -> OpResult Node.AllocElement
474 allocateOnPair nl inst new_pdx new_sdx =
475 let tgt_p = Container.find new_pdx nl
476 tgt_s = Container.find new_sdx nl
478 Instance.instMatchesPolicy inst (Node.iPolicy tgt_p)
479 new_p <- Node.addPri tgt_p inst
480 new_s <- Node.addSec tgt_s inst new_pdx
481 let new_inst = Instance.setBoth inst new_pdx new_sdx
482 new_nl = Container.addTwo new_pdx new_p new_sdx new_s nl
483 return (new_nl, new_inst, [new_p, new_s], compCV new_nl)
485 -- | Tries to perform an instance move and returns the best table
486 -- between the original one and the new one.
487 checkSingleStep :: Table -- ^ The original table
488 -> Instance.Instance -- ^ The instance to move
489 -> Table -- ^ The current best table
490 -> IMove -- ^ The move to apply
491 -> Table -- ^ The final best table
492 checkSingleStep ini_tbl target cur_tbl move =
493 let Table ini_nl ini_il _ ini_plc = ini_tbl
494 tmp_resu = applyMove ini_nl target move
497 OpGood (upd_nl, new_inst, pri_idx, sec_idx) ->
498 let tgt_idx = Instance.idx target
499 upd_cvar = compCV upd_nl
500 upd_il = Container.add tgt_idx new_inst ini_il
501 upd_plc = (tgt_idx, pri_idx, sec_idx, move, upd_cvar):ini_plc
502 upd_tbl = Table upd_nl upd_il upd_cvar upd_plc
503 in compareTables cur_tbl upd_tbl
505 -- | Given the status of the current secondary as a valid new node and
506 -- the current candidate target node, generate the possible moves for
508 possibleMoves :: Bool -- ^ Whether the secondary node is a valid new node
509 -> Bool -- ^ Whether we can change the primary node
510 -> Ndx -- ^ Target node candidate
511 -> [IMove] -- ^ List of valid result moves
513 possibleMoves _ False tdx =
514 [ReplaceSecondary tdx]
516 possibleMoves True True tdx =
517 [ ReplaceSecondary tdx
518 , ReplaceAndFailover tdx
520 , FailoverAndReplace tdx
523 possibleMoves False True tdx =
524 [ ReplaceSecondary tdx
525 , ReplaceAndFailover tdx
528 -- | Compute the best move for a given instance.
529 checkInstanceMove :: [Ndx] -- ^ Allowed target node indices
530 -> Bool -- ^ Whether disk moves are allowed
531 -> Bool -- ^ Whether instance moves are allowed
532 -> Table -- ^ Original table
533 -> Instance.Instance -- ^ Instance to move
534 -> Table -- ^ Best new table for this instance
535 checkInstanceMove nodes_idx disk_moves inst_moves ini_tbl target =
536 let opdx = Instance.pNode target
537 osdx = Instance.sNode target
538 bad_nodes = [opdx, osdx]
539 nodes = filter (`notElem` bad_nodes) nodes_idx
540 use_secondary = elem osdx nodes_idx && inst_moves
541 aft_failover = if use_secondary -- if allowed to failover
542 then checkSingleStep ini_tbl target ini_tbl Failover
544 all_moves = if disk_moves
546 (possibleMoves use_secondary inst_moves) nodes
549 -- iterate over the possible nodes for this instance
550 foldl' (checkSingleStep ini_tbl target) aft_failover all_moves
552 -- | Compute the best next move.
553 checkMove :: [Ndx] -- ^ Allowed target node indices
554 -> Bool -- ^ Whether disk moves are allowed
555 -> Bool -- ^ Whether instance moves are allowed
556 -> Table -- ^ The current solution
557 -> [Instance.Instance] -- ^ List of instances still to move
558 -> Table -- ^ The new solution
559 checkMove nodes_idx disk_moves inst_moves ini_tbl victims =
560 let Table _ _ _ ini_plc = ini_tbl
561 -- we're using rwhnf from the Control.Parallel.Strategies
562 -- package; we don't need to use rnf as that would force too
563 -- much evaluation in single-threaded cases, and in
564 -- multi-threaded case the weak head normal form is enough to
565 -- spark the evaluation
566 tables = parMap rwhnf (checkInstanceMove nodes_idx disk_moves
569 -- iterate over all instances, computing the best move
570 best_tbl = foldl' compareTables ini_tbl tables
571 Table _ _ _ best_plc = best_tbl
572 in if length best_plc == length ini_plc
573 then ini_tbl -- no advancement
576 -- | Check if we are allowed to go deeper in the balancing.
577 doNextBalance :: Table -- ^ The starting table
578 -> Int -- ^ Remaining length
579 -> Score -- ^ Score at which to stop
580 -> Bool -- ^ The resulting table and commands
581 doNextBalance ini_tbl max_rounds min_score =
582 let Table _ _ ini_cv ini_plc = ini_tbl
583 ini_plc_len = length ini_plc
584 in (max_rounds < 0 || ini_plc_len < max_rounds) && ini_cv > min_score
586 -- | Run a balance move.
587 tryBalance :: Table -- ^ The starting table
588 -> Bool -- ^ Allow disk moves
589 -> Bool -- ^ Allow instance moves
590 -> Bool -- ^ Only evacuate moves
591 -> Score -- ^ Min gain threshold
592 -> Score -- ^ Min gain
593 -> Maybe Table -- ^ The resulting table and commands
594 tryBalance ini_tbl disk_moves inst_moves evac_mode mg_limit min_gain =
595 let Table ini_nl ini_il ini_cv _ = ini_tbl
596 all_inst = Container.elems ini_il
597 all_inst' = if evac_mode
598 then let bad_nodes = map Node.idx . filter Node.offline $
599 Container.elems ini_nl
600 in filter (any (`elem` bad_nodes) . Instance.allNodes)
603 reloc_inst = filter Instance.movable all_inst'
604 node_idx = map Node.idx . filter (not . Node.offline) $
605 Container.elems ini_nl
606 fin_tbl = checkMove node_idx disk_moves inst_moves ini_tbl reloc_inst
607 (Table _ _ fin_cv _) = fin_tbl
609 if fin_cv < ini_cv && (ini_cv > mg_limit || ini_cv - fin_cv >= min_gain)
610 then Just fin_tbl -- this round made success, return the new table
613 -- * Allocation functions
615 -- | Build failure stats out of a list of failures.
616 collapseFailures :: [FailMode] -> FailStats
617 collapseFailures flst =
618 map (\k -> (k, foldl' (\a e -> if e == k then a + 1 else a) 0 flst))
621 -- | Compares two Maybe AllocElement and chooses the besst score.
622 bestAllocElement :: Maybe Node.AllocElement
623 -> Maybe Node.AllocElement
624 -> Maybe Node.AllocElement
625 bestAllocElement a Nothing = a
626 bestAllocElement Nothing b = b
627 bestAllocElement a@(Just (_, _, _, ascore)) b@(Just (_, _, _, bscore)) =
628 if ascore < bscore then a else b
630 -- | Update current Allocation solution and failure stats with new
632 concatAllocs :: AllocSolution -> OpResult Node.AllocElement -> AllocSolution
633 concatAllocs as (OpFail reason) = as { asFailures = reason : asFailures as }
635 concatAllocs as (OpGood ns) =
636 let -- Choose the old or new solution, based on the cluster score
638 osols = asSolution as
639 nsols = bestAllocElement osols (Just ns)
641 -- Note: we force evaluation of nsols here in order to keep the
642 -- memory profile low - we know that we will need nsols for sure
643 -- in the next cycle, so we force evaluation of nsols, since the
644 -- foldl' in the caller will only evaluate the tuple, but not the
645 -- elements of the tuple
646 in nsols `seq` nsuc `seq` as { asAllocs = nsuc, asSolution = nsols }
648 -- | Sums two 'AllocSolution' structures.
649 sumAllocs :: AllocSolution -> AllocSolution -> AllocSolution
650 sumAllocs (AllocSolution aFails aAllocs aSols aLog)
651 (AllocSolution bFails bAllocs bSols bLog) =
652 -- note: we add b first, since usually it will be smaller; when
653 -- fold'ing, a will grow and grow whereas b is the per-group
654 -- result, hence smaller
655 let nFails = bFails ++ aFails
656 nAllocs = aAllocs + bAllocs
657 nSols = bestAllocElement aSols bSols
659 in AllocSolution nFails nAllocs nSols nLog
661 -- | Given a solution, generates a reasonable description for it.
662 describeSolution :: AllocSolution -> String
663 describeSolution as =
664 let fcnt = asFailures as
667 intercalate ", " . map (\(a, b) -> printf "%s: %d" (show a) b) .
668 filter ((> 0) . snd) . collapseFailures $ fcnt
670 Nothing -> "No valid allocation solutions, failure reasons: " ++
671 (if null fcnt then "unknown reasons" else freasons)
672 Just (_, _, nodes, cv) ->
673 printf ("score: %.8f, successes %d, failures %d (%s)" ++
674 " for node(s) %s") cv (asAllocs as) (length fcnt) freasons
675 (intercalate "/" . map Node.name $ nodes)
677 -- | Annotates a solution with the appropriate string.
678 annotateSolution :: AllocSolution -> AllocSolution
679 annotateSolution as = as { asLog = describeSolution as : asLog as }
681 -- | Reverses an evacuation solution.
683 -- Rationale: we always concat the results to the top of the lists, so
684 -- for proper jobset execution, we should reverse all lists.
685 reverseEvacSolution :: EvacSolution -> EvacSolution
686 reverseEvacSolution (EvacSolution f m o) =
687 EvacSolution (reverse f) (reverse m) (reverse o)
689 -- | Generate the valid node allocation singles or pairs for a new instance.
690 genAllocNodes :: Group.List -- ^ Group list
691 -> Node.List -- ^ The node map
692 -> Int -- ^ The number of nodes required
693 -> Bool -- ^ Whether to drop or not
695 -> Result AllocNodes -- ^ The (monadic) result
696 genAllocNodes gl nl count drop_unalloc =
697 let filter_fn = if drop_unalloc
698 then filter (Group.isAllocable .
699 flip Container.find gl . Node.group)
701 all_nodes = filter_fn $ getOnline nl
702 all_pairs = [(Node.idx p,
703 [Node.idx s | s <- all_nodes,
704 Node.idx p /= Node.idx s,
705 Node.group p == Node.group s]) |
708 1 -> Ok (Left (map Node.idx all_nodes))
709 2 -> Ok (Right (filter (not . null . snd) all_pairs))
710 _ -> Bad "Unsupported number of nodes, only one or two supported"
712 -- | Try to allocate an instance on the cluster.
713 tryAlloc :: (Monad m) =>
714 Node.List -- ^ The node list
715 -> Instance.List -- ^ The instance list
716 -> Instance.Instance -- ^ The instance to allocate
717 -> AllocNodes -- ^ The allocation targets
718 -> m AllocSolution -- ^ Possible solution list
719 tryAlloc _ _ _ (Right []) = fail "Not enough online nodes"
720 tryAlloc nl _ inst (Right ok_pairs) =
721 let psols = parMap rwhnf (\(p, ss) ->
723 concatAllocs cstate .
724 allocateOnPair nl inst p)
725 emptyAllocSolution ss) ok_pairs
726 sols = foldl' sumAllocs emptyAllocSolution psols
727 in return $ annotateSolution sols
729 tryAlloc _ _ _ (Left []) = fail "No online nodes"
730 tryAlloc nl _ inst (Left all_nodes) =
731 let sols = foldl' (\cstate ->
732 concatAllocs cstate . allocateOnSingle nl inst
733 ) emptyAllocSolution all_nodes
734 in return $ annotateSolution sols
736 -- | Given a group/result, describe it as a nice (list of) messages.
737 solutionDescription :: Group.List -> (Gdx, Result AllocSolution) -> [String]
738 solutionDescription gl (groupId, result) =
740 Ok solution -> map (printf "Group %s (%s): %s" gname pol) (asLog solution)
741 Bad message -> [printf "Group %s: error %s" gname message]
742 where grp = Container.find groupId gl
743 gname = Group.name grp
744 pol = allocPolicyToRaw (Group.allocPolicy grp)
746 -- | From a list of possibly bad and possibly empty solutions, filter
747 -- only the groups with a valid result. Note that the result will be
748 -- reversed compared to the original list.
749 filterMGResults :: Group.List
750 -> [(Gdx, Result AllocSolution)]
751 -> [(Gdx, AllocSolution)]
752 filterMGResults gl = foldl' fn []
753 where unallocable = not . Group.isAllocable . flip Container.find gl
754 fn accu (gdx, rasol) =
757 Ok sol | isNothing (asSolution sol) -> accu
758 | unallocable gdx -> accu
759 | otherwise -> (gdx, sol):accu
761 -- | Sort multigroup results based on policy and score.
762 sortMGResults :: Group.List
763 -> [(Gdx, AllocSolution)]
764 -> [(Gdx, AllocSolution)]
765 sortMGResults gl sols =
766 let extractScore (_, _, _, x) = x
767 solScore (gdx, sol) = (Group.allocPolicy (Container.find gdx gl),
768 (extractScore . fromJust . asSolution) sol)
769 in sortBy (comparing solScore) sols
771 -- | Finds the best group for an instance on a multi-group cluster.
773 -- Only solutions in @preferred@ and @last_resort@ groups will be
774 -- accepted as valid, and additionally if the allowed groups parameter
775 -- is not null then allocation will only be run for those group
777 findBestAllocGroup :: Group.List -- ^ The group list
778 -> Node.List -- ^ The node list
779 -> Instance.List -- ^ The instance list
780 -> Maybe [Gdx] -- ^ The allowed groups
781 -> Instance.Instance -- ^ The instance to allocate
782 -> Int -- ^ Required number of nodes
783 -> Result (Gdx, AllocSolution, [String])
784 findBestAllocGroup mggl mgnl mgil allowed_gdxs inst cnt =
785 let groups = splitCluster mgnl mgil
786 groups' = maybe groups (\gs -> filter ((`elem` gs) . fst) groups)
788 sols = map (\(gid, (nl, il)) ->
789 (gid, genAllocNodes mggl nl cnt False >>=
790 tryAlloc nl il inst))
791 groups'::[(Gdx, Result AllocSolution)]
792 all_msgs = concatMap (solutionDescription mggl) sols
793 goodSols = filterMGResults mggl sols
794 sortedSols = sortMGResults mggl goodSols
795 in if null sortedSols
797 then Bad $ "no groups for evacuation: allowed groups was" ++
798 show allowed_gdxs ++ ", all groups: " ++
799 show (map fst groups)
800 else Bad $ intercalate ", " all_msgs
801 else let (final_group, final_sol) = head sortedSols
802 in return (final_group, final_sol, all_msgs)
804 -- | Try to allocate an instance on a multi-group cluster.
805 tryMGAlloc :: Group.List -- ^ The group list
806 -> Node.List -- ^ The node list
807 -> Instance.List -- ^ The instance list
808 -> Instance.Instance -- ^ The instance to allocate
809 -> Int -- ^ Required number of nodes
810 -> Result AllocSolution -- ^ Possible solution list
811 tryMGAlloc mggl mgnl mgil inst cnt = do
812 (best_group, solution, all_msgs) <-
813 findBestAllocGroup mggl mgnl mgil Nothing inst cnt
814 let group_name = Group.name $ Container.find best_group mggl
815 selmsg = "Selected group: " ++ group_name
816 return $ solution { asLog = selmsg:all_msgs }
818 -- | Function which fails if the requested mode is change secondary.
820 -- This is useful since except DRBD, no other disk template can
821 -- execute change secondary; thus, we can just call this function
822 -- instead of always checking for secondary mode. After the call to
823 -- this function, whatever mode we have is just a primary change.
824 failOnSecondaryChange :: (Monad m) => EvacMode -> DiskTemplate -> m ()
825 failOnSecondaryChange ChangeSecondary dt =
826 fail $ "Instances with disk template '" ++ diskTemplateToRaw dt ++
827 "' can't execute change secondary"
828 failOnSecondaryChange _ _ = return ()
830 -- | Run evacuation for a single instance.
832 -- /Note:/ this function should correctly execute both intra-group
833 -- evacuations (in all modes) and inter-group evacuations (in the
834 -- 'ChangeAll' mode). Of course, this requires that the correct list
835 -- of target nodes is passed.
836 nodeEvacInstance :: Node.List -- ^ The node list (cluster-wide)
837 -> Instance.List -- ^ Instance list (cluster-wide)
838 -> EvacMode -- ^ The evacuation mode
839 -> Instance.Instance -- ^ The instance to be evacuated
840 -> Gdx -- ^ The group we're targetting
841 -> [Ndx] -- ^ The list of available nodes
843 -> Result (Node.List, Instance.List, [OpCodes.OpCode])
844 nodeEvacInstance _ _ mode (Instance.Instance
845 {Instance.diskTemplate = dt@DTDiskless}) _ _ =
846 failOnSecondaryChange mode dt >>
847 fail "Diskless relocations not implemented yet"
849 nodeEvacInstance _ _ _ (Instance.Instance
850 {Instance.diskTemplate = DTPlain}) _ _ =
851 fail "Instances of type plain cannot be relocated"
853 nodeEvacInstance _ _ _ (Instance.Instance
854 {Instance.diskTemplate = DTFile}) _ _ =
855 fail "Instances of type file cannot be relocated"
857 nodeEvacInstance _ _ mode (Instance.Instance
858 {Instance.diskTemplate = dt@DTSharedFile}) _ _ =
859 failOnSecondaryChange mode dt >>
860 fail "Shared file relocations not implemented yet"
862 nodeEvacInstance _ _ mode (Instance.Instance
863 {Instance.diskTemplate = dt@DTBlock}) _ _ =
864 failOnSecondaryChange mode dt >>
865 fail "Block device relocations not implemented yet"
867 nodeEvacInstance _ _ mode (Instance.Instance
868 {Instance.diskTemplate = dt@DTRbd}) _ _ =
869 failOnSecondaryChange mode dt >>
870 fail "Rbd relocations not implemented yet"
872 nodeEvacInstance nl il ChangePrimary
873 inst@(Instance.Instance {Instance.diskTemplate = DTDrbd8})
876 (nl', inst', _, _) <- opToResult $ applyMove nl inst Failover
877 let idx = Instance.idx inst
878 il' = Container.add idx inst' il
879 ops = iMoveToJob nl' il' idx Failover
880 return (nl', il', ops)
882 nodeEvacInstance nl il ChangeSecondary
883 inst@(Instance.Instance {Instance.diskTemplate = DTDrbd8})
886 (nl', inst', _, ndx) <- annotateResult "Can't find any good node" $
888 foldl' (evacDrbdSecondaryInner nl inst gdx)
889 (Left "no nodes available") avail_nodes
890 let idx = Instance.idx inst
891 il' = Container.add idx inst' il
892 ops = iMoveToJob nl' il' idx (ReplaceSecondary ndx)
893 return (nl', il', ops)
895 -- The algorithm for ChangeAll is as follows:
897 -- * generate all (primary, secondary) node pairs for the target groups
898 -- * for each pair, execute the needed moves (r:s, f, r:s) and compute
899 -- the final node list state and group score
900 -- * select the best choice via a foldl that uses the same Either
901 -- String solution as the ChangeSecondary mode
902 nodeEvacInstance nl il ChangeAll
903 inst@(Instance.Instance {Instance.diskTemplate = DTDrbd8})
906 let no_nodes = Left "no nodes available"
907 node_pairs = [(p,s) | p <- avail_nodes, s <- avail_nodes, p /= s]
908 (nl', il', ops, _) <-
909 annotateResult "Can't find any good nodes for relocation" $
912 (\accu nodes -> case evacDrbdAllInner nl il inst gdx nodes of
916 -- we don't need more details (which
917 -- nodes, etc.) as we only selected
918 -- this group if we can allocate on
919 -- it, hence failures will not
920 -- propagate out of this fold loop
921 Left _ -> Left $ "Allocation failed: " ++ msg
922 Ok result@(_, _, _, new_cv) ->
923 let new_accu = Right result in
926 Right (_, _, _, old_cv) ->
930 ) no_nodes node_pairs
932 return (nl', il', ops)
934 -- | Inner fold function for changing secondary of a DRBD instance.
936 -- The running solution is either a @Left String@, which means we
937 -- don't have yet a working solution, or a @Right (...)@, which
938 -- represents a valid solution; it holds the modified node list, the
939 -- modified instance (after evacuation), the score of that solution,
940 -- and the new secondary node index.
941 evacDrbdSecondaryInner :: Node.List -- ^ Cluster node list
942 -> Instance.Instance -- ^ Instance being evacuated
943 -> Gdx -- ^ The group index of the instance
944 -> Either String ( Node.List
947 , Ndx) -- ^ Current best solution
948 -> Ndx -- ^ Node we're evaluating as new secondary
949 -> Either String ( Node.List
952 , Ndx) -- ^ New best solution
953 evacDrbdSecondaryInner nl inst gdx accu ndx =
954 case applyMove nl inst (ReplaceSecondary ndx) of
958 Left _ -> Left $ "Node " ++ Container.nameOf nl ndx ++
959 " failed: " ++ show fm
960 OpGood (nl', inst', _, _) ->
961 let nodes = Container.elems nl'
962 -- The fromJust below is ugly (it can fail nastily), but
963 -- at this point we should have any internal mismatches,
964 -- and adding a monad here would be quite involved
965 grpnodes = fromJust (gdx `lookup` Node.computeGroups nodes)
966 new_cv = compCVNodes grpnodes
967 new_accu = Right (nl', inst', new_cv, ndx)
970 Right (_, _, old_cv, _) ->
975 -- | Compute result of changing all nodes of a DRBD instance.
977 -- Given the target primary and secondary node (which might be in a
978 -- different group or not), this function will 'execute' all the
979 -- required steps and assuming all operations succceed, will return
980 -- the modified node and instance lists, the opcodes needed for this
981 -- and the new group score.
982 evacDrbdAllInner :: Node.List -- ^ Cluster node list
983 -> Instance.List -- ^ Cluster instance list
984 -> Instance.Instance -- ^ The instance to be moved
985 -> Gdx -- ^ The target group index
986 -- (which can differ from the
987 -- current group of the
989 -> (Ndx, Ndx) -- ^ Tuple of new
990 -- primary\/secondary nodes
991 -> Result (Node.List, Instance.List, [OpCodes.OpCode], Score)
992 evacDrbdAllInner nl il inst gdx (t_pdx, t_sdx) = do
993 let primary = Container.find (Instance.pNode inst) nl
994 idx = Instance.idx inst
995 -- if the primary is offline, then we first failover
996 (nl1, inst1, ops1) <-
997 if Node.offline primary
999 (nl', inst', _, _) <-
1000 annotateResult "Failing over to the secondary" $
1001 opToResult $ applyMove nl inst Failover
1002 return (nl', inst', [Failover])
1003 else return (nl, inst, [])
1004 let (o1, o2, o3) = (ReplaceSecondary t_pdx,
1006 ReplaceSecondary t_sdx)
1007 -- we now need to execute a replace secondary to the future
1009 (nl2, inst2, _, _) <-
1010 annotateResult "Changing secondary to new primary" $
1012 applyMove nl1 inst1 o1
1014 -- we now execute another failover, the primary stays fixed now
1015 (nl3, inst3, _, _) <- annotateResult "Failing over to new primary" $
1016 opToResult $ applyMove nl2 inst2 o2
1018 -- and finally another replace secondary, to the final secondary
1019 (nl4, inst4, _, _) <-
1020 annotateResult "Changing secondary to final secondary" $
1022 applyMove nl3 inst3 o3
1024 il' = Container.add idx inst4 il
1025 ops = concatMap (iMoveToJob nl4 il' idx) $ reverse ops4
1026 let nodes = Container.elems nl4
1027 -- The fromJust below is ugly (it can fail nastily), but
1028 -- at this point we should have any internal mismatches,
1029 -- and adding a monad here would be quite involved
1030 grpnodes = fromJust (gdx `lookup` Node.computeGroups nodes)
1031 new_cv = compCVNodes grpnodes
1032 return (nl4, il', ops, new_cv)
1034 -- | Computes the nodes in a given group which are available for
1036 availableGroupNodes :: [(Gdx, [Ndx])] -- ^ Group index/node index assoc list
1037 -> IntSet.IntSet -- ^ Nodes that are excluded
1038 -> Gdx -- ^ The group for which we
1040 -> Result [Ndx] -- ^ List of available node indices
1041 availableGroupNodes group_nodes excl_ndx gdx = do
1042 local_nodes <- maybe (Bad $ "Can't find group with index " ++ show gdx)
1043 Ok (lookup gdx group_nodes)
1044 let avail_nodes = filter (not . flip IntSet.member excl_ndx) local_nodes
1047 -- | Updates the evac solution with the results of an instance
1049 updateEvacSolution :: (Node.List, Instance.List, EvacSolution)
1051 -> Result (Node.List, Instance.List, [OpCodes.OpCode])
1052 -> (Node.List, Instance.List, EvacSolution)
1053 updateEvacSolution (nl, il, es) idx (Bad msg) =
1054 (nl, il, es { esFailed = (idx, msg):esFailed es})
1055 updateEvacSolution (_, _, es) idx (Ok (nl, il, opcodes)) =
1056 (nl, il, es { esMoved = new_elem:esMoved es
1057 , esOpCodes = opcodes:esOpCodes es })
1058 where inst = Container.find idx il
1060 instancePriGroup nl inst,
1061 Instance.allNodes inst)
1063 -- | Node-evacuation IAllocator mode main function.
1064 tryNodeEvac :: Group.List -- ^ The cluster groups
1065 -> Node.List -- ^ The node list (cluster-wide, not per group)
1066 -> Instance.List -- ^ Instance list (cluster-wide)
1067 -> EvacMode -- ^ The evacuation mode
1068 -> [Idx] -- ^ List of instance (indices) to be evacuated
1069 -> Result (Node.List, Instance.List, EvacSolution)
1070 tryNodeEvac _ ini_nl ini_il mode idxs =
1071 let evac_ndx = nodesToEvacuate ini_il mode idxs
1072 offline = map Node.idx . filter Node.offline $ Container.elems ini_nl
1073 excl_ndx = foldl' (flip IntSet.insert) evac_ndx offline
1074 group_ndx = map (\(gdx, (nl, _)) -> (gdx, map Node.idx
1075 (Container.elems nl))) $
1076 splitCluster ini_nl ini_il
1077 (fin_nl, fin_il, esol) =
1078 foldl' (\state@(nl, il, _) inst ->
1079 let gdx = instancePriGroup nl inst
1080 pdx = Instance.pNode inst in
1081 updateEvacSolution state (Instance.idx inst) $
1082 availableGroupNodes group_ndx
1083 (IntSet.insert pdx excl_ndx) gdx >>=
1084 nodeEvacInstance nl il mode inst gdx
1086 (ini_nl, ini_il, emptyEvacSolution)
1087 (map (`Container.find` ini_il) idxs)
1088 in return (fin_nl, fin_il, reverseEvacSolution esol)
1090 -- | Change-group IAllocator mode main function.
1092 -- This is very similar to 'tryNodeEvac', the only difference is that
1093 -- we don't choose as target group the current instance group, but
1096 -- 1. at the start of the function, we compute which are the target
1097 -- groups; either no groups were passed in, in which case we choose
1098 -- all groups out of which we don't evacuate instance, or there were
1099 -- some groups passed, in which case we use those
1101 -- 2. for each instance, we use 'findBestAllocGroup' to choose the
1102 -- best group to hold the instance, and then we do what
1103 -- 'tryNodeEvac' does, except for this group instead of the current
1106 -- Note that the correct behaviour of this function relies on the
1107 -- function 'nodeEvacInstance' to be able to do correctly both
1108 -- intra-group and inter-group moves when passed the 'ChangeAll' mode.
1109 tryChangeGroup :: Group.List -- ^ The cluster groups
1110 -> Node.List -- ^ The node list (cluster-wide)
1111 -> Instance.List -- ^ Instance list (cluster-wide)
1112 -> [Gdx] -- ^ Target groups; if empty, any
1113 -- groups not being evacuated
1114 -> [Idx] -- ^ List of instance (indices) to be evacuated
1115 -> Result (Node.List, Instance.List, EvacSolution)
1116 tryChangeGroup gl ini_nl ini_il gdxs idxs =
1117 let evac_gdxs = nub $ map (instancePriGroup ini_nl .
1118 flip Container.find ini_il) idxs
1119 target_gdxs = (if null gdxs
1120 then Container.keys gl
1121 else gdxs) \\ evac_gdxs
1122 offline = map Node.idx . filter Node.offline $ Container.elems ini_nl
1123 excl_ndx = foldl' (flip IntSet.insert) IntSet.empty offline
1124 group_ndx = map (\(gdx, (nl, _)) -> (gdx, map Node.idx
1125 (Container.elems nl))) $
1126 splitCluster ini_nl ini_il
1127 (fin_nl, fin_il, esol) =
1128 foldl' (\state@(nl, il, _) inst ->
1130 let ncnt = Instance.requiredNodes $
1131 Instance.diskTemplate inst
1132 (gdx, _, _) <- findBestAllocGroup gl nl il
1133 (Just target_gdxs) inst ncnt
1134 av_nodes <- availableGroupNodes group_ndx
1136 nodeEvacInstance nl il ChangeAll inst gdx av_nodes
1137 in updateEvacSolution state (Instance.idx inst) solution
1139 (ini_nl, ini_il, emptyEvacSolution)
1140 (map (`Container.find` ini_il) idxs)
1141 in return (fin_nl, fin_il, reverseEvacSolution esol)
1143 -- | Standard-sized allocation method.
1145 -- This places instances of the same size on the cluster until we're
1146 -- out of space. The result will be a list of identically-sized
1148 iterateAlloc :: AllocMethod
1149 iterateAlloc nl il limit newinst allocnodes ixes cstats =
1150 let depth = length ixes
1151 newname = printf "new-%d" depth::String
1152 newidx = Container.size il
1153 newi2 = Instance.setIdx (Instance.setName newinst newname) newidx
1154 newlimit = fmap (flip (-) 1) limit
1155 in case tryAlloc nl il newi2 allocnodes of
1157 Ok (AllocSolution { asFailures = errs, asSolution = sols3 }) ->
1158 let newsol = Ok (collapseFailures errs, nl, il, ixes, cstats) in
1161 Just (xnl, xi, _, _) ->
1164 else iterateAlloc xnl (Container.add newidx xi il)
1165 newlimit newinst allocnodes (xi:ixes)
1166 (totalResources xnl:cstats)
1168 -- | Tiered allocation method.
1170 -- This places instances on the cluster, and decreases the spec until
1171 -- we can allocate again. The result will be a list of decreasing
1173 tieredAlloc :: AllocMethod
1174 tieredAlloc nl il limit newinst allocnodes ixes cstats =
1175 case iterateAlloc nl il limit newinst allocnodes ixes cstats of
1177 Ok (errs, nl', il', ixes', cstats') ->
1178 let newsol = Ok (errs, nl', il', ixes', cstats')
1179 ixes_cnt = length ixes'
1180 (stop, newlimit) = case limit of
1181 Nothing -> (False, Nothing)
1182 Just n -> (n <= ixes_cnt,
1183 Just (n - ixes_cnt)) in
1184 if stop then newsol else
1185 case Instance.shrinkByType newinst . fst . last $
1186 sortBy (comparing snd) errs of
1188 Ok newinst' -> tieredAlloc nl' il' newlimit
1189 newinst' allocnodes ixes' cstats'
1191 -- * Formatting functions
1193 -- | Given the original and final nodes, computes the relocation description.
1194 computeMoves :: Instance.Instance -- ^ The instance to be moved
1195 -> String -- ^ The instance name
1196 -> IMove -- ^ The move being performed
1197 -> String -- ^ New primary
1198 -> String -- ^ New secondary
1199 -> (String, [String])
1200 -- ^ Tuple of moves and commands list; moves is containing
1201 -- either @/f/@ for failover or @/r:name/@ for replace
1202 -- secondary, while the command list holds gnt-instance
1203 -- commands (without that prefix), e.g \"@failover instance1@\"
1204 computeMoves i inam mv c d =
1206 Failover -> ("f", [mig])
1207 FailoverAndReplace _ -> (printf "f r:%s" d, [mig, rep d])
1208 ReplaceSecondary _ -> (printf "r:%s" d, [rep d])
1209 ReplaceAndFailover _ -> (printf "r:%s f" c, [rep c, mig])
1210 ReplacePrimary _ -> (printf "f r:%s f" c, [mig, rep c, mig])
1211 where morf = if Instance.instanceRunning i then "migrate" else "failover"
1212 mig = printf "%s -f %s" morf inam::String
1213 rep n = printf "replace-disks -n %s %s" n inam
1215 -- | Converts a placement to string format.
1216 printSolutionLine :: Node.List -- ^ The node list
1217 -> Instance.List -- ^ The instance list
1218 -> Int -- ^ Maximum node name length
1219 -> Int -- ^ Maximum instance name length
1220 -> Placement -- ^ The current placement
1221 -> Int -- ^ The index of the placement in
1223 -> (String, [String])
1224 printSolutionLine nl il nmlen imlen plc pos =
1225 let pmlen = (2*nmlen + 1)
1226 (i, p, s, mv, c) = plc
1227 inst = Container.find i il
1228 inam = Instance.alias inst
1229 npri = Node.alias $ Container.find p nl
1230 nsec = Node.alias $ Container.find s nl
1231 opri = Node.alias $ Container.find (Instance.pNode inst) nl
1232 osec = Node.alias $ Container.find (Instance.sNode inst) nl
1233 (moves, cmds) = computeMoves inst inam mv npri nsec
1234 ostr = printf "%s:%s" opri osec::String
1235 nstr = printf "%s:%s" npri nsec::String
1236 in (printf " %3d. %-*s %-*s => %-*s %.8f a=%s"
1237 pos imlen inam pmlen ostr
1241 -- | Return the instance and involved nodes in an instance move.
1243 -- Note that the output list length can vary, and is not required nor
1244 -- guaranteed to be of any specific length.
1245 involvedNodes :: Instance.List -- ^ Instance list, used for retrieving
1246 -- the instance from its index; note
1247 -- that this /must/ be the original
1248 -- instance list, so that we can
1249 -- retrieve the old nodes
1250 -> Placement -- ^ The placement we're investigating,
1251 -- containing the new nodes and
1253 -> [Ndx] -- ^ Resulting list of node indices
1254 involvedNodes il plc =
1255 let (i, np, ns, _, _) = plc
1256 inst = Container.find i il
1257 in nub $ [np, ns] ++ Instance.allNodes inst
1259 -- | Inner function for splitJobs, that either appends the next job to
1260 -- the current jobset, or starts a new jobset.
1261 mergeJobs :: ([JobSet], [Ndx]) -> MoveJob -> ([JobSet], [Ndx])
1262 mergeJobs ([], _) n@(ndx, _, _, _) = ([[n]], ndx)
1263 mergeJobs (cjs@(j:js), nbuf) n@(ndx, _, _, _)
1264 | null (ndx `intersect` nbuf) = ((n:j):js, ndx ++ nbuf)
1265 | otherwise = ([n]:cjs, ndx)
1267 -- | Break a list of moves into independent groups. Note that this
1268 -- will reverse the order of jobs.
1269 splitJobs :: [MoveJob] -> [JobSet]
1270 splitJobs = fst . foldl mergeJobs ([], [])
1272 -- | Given a list of commands, prefix them with @gnt-instance@ and
1273 -- also beautify the display a little.
1274 formatJob :: Int -> Int -> (Int, MoveJob) -> [String]
1275 formatJob jsn jsl (sn, (_, _, _, cmds)) =
1277 printf " echo job %d/%d" jsn sn:
1279 map (" gnt-instance " ++) cmds
1281 then ["", printf "echo jobset %d, %d jobs" jsn jsl] ++ out
1284 -- | Given a list of commands, prefix them with @gnt-instance@ and
1285 -- also beautify the display a little.
1286 formatCmds :: [JobSet] -> String
1289 concatMap (\(jsn, js) -> concatMap (formatJob jsn (length js))
1293 -- | Print the node list.
1294 printNodes :: Node.List -> [String] -> String
1296 let fields = case fs of
1297 [] -> Node.defaultFields
1298 "+":rest -> Node.defaultFields ++ rest
1300 snl = sortBy (comparing Node.idx) (Container.elems nl)
1301 (header, isnum) = unzip $ map Node.showHeader fields
1302 in unlines . map ((:) ' ' . unwords) $
1303 formatTable (header:map (Node.list fields) snl) isnum
1305 -- | Print the instance list.
1306 printInsts :: Node.List -> Instance.List -> String
1308 let sil = sortBy (comparing Instance.idx) (Container.elems il)
1309 helper inst = [ if Instance.instanceRunning inst then "R" else " "
1310 , Instance.name inst
1311 , Container.nameOf nl (Instance.pNode inst)
1312 , let sdx = Instance.sNode inst
1313 in if sdx == Node.noSecondary
1315 else Container.nameOf nl sdx
1316 , if Instance.autoBalance inst then "Y" else "N"
1317 , printf "%3d" $ Instance.vcpus inst
1318 , printf "%5d" $ Instance.mem inst
1319 , printf "%5d" $ Instance.dsk inst `div` 1024
1325 where DynUtil lC lM lD lN = Instance.util inst
1326 header = [ "F", "Name", "Pri_node", "Sec_node", "Auto_bal"
1327 , "vcpu", "mem" , "dsk", "lCpu", "lMem", "lDsk", "lNet" ]
1328 isnum = False:False:False:False:False:repeat True
1329 in unlines . map ((:) ' ' . unwords) $
1330 formatTable (header:map helper sil) isnum
1332 -- | Shows statistics for a given node list.
1333 printStats :: String -> Node.List -> String
1335 let dcvs = compDetailedCV $ Container.elems nl
1336 (weights, names) = unzip detailedCVInfo
1337 hd = zip3 (weights ++ repeat 1) (names ++ repeat "unknown") dcvs
1338 header = [ "Field", "Value", "Weight" ]
1339 formatted = map (\(w, h, val) ->
1344 in unlines . map ((++) lp) . map ((:) ' ' . unwords) $
1345 formatTable (header:formatted) $ False:repeat True
1347 -- | Convert a placement into a list of OpCodes (basically a job).
1348 iMoveToJob :: Node.List -- ^ The node list; only used for node
1349 -- names, so any version is good
1350 -- (before or after the operation)
1351 -> Instance.List -- ^ The instance list; also used for
1353 -> Idx -- ^ The index of the instance being
1355 -> IMove -- ^ The actual move to be described
1356 -> [OpCodes.OpCode] -- ^ The list of opcodes equivalent to
1358 iMoveToJob nl il idx move =
1359 let inst = Container.find idx il
1360 iname = Instance.name inst
1361 lookNode = Just . Container.nameOf nl
1362 opF = OpCodes.OpInstanceMigrate iname True False True Nothing
1363 opR n = OpCodes.OpInstanceReplaceDisks iname (lookNode n)
1364 OpCodes.ReplaceNewSecondary [] Nothing
1367 ReplacePrimary np -> [ opF, opR np, opF ]
1368 ReplaceSecondary ns -> [ opR ns ]
1369 ReplaceAndFailover np -> [ opR np, opF ]
1370 FailoverAndReplace ns -> [ opF, opR ns ]
1372 -- * Node group functions
1374 -- | Computes the group of an instance.
1375 instanceGroup :: Node.List -> Instance.Instance -> Result Gdx
1376 instanceGroup nl i =
1377 let sidx = Instance.sNode i
1378 pnode = Container.find (Instance.pNode i) nl
1379 snode = if sidx == Node.noSecondary
1381 else Container.find sidx nl
1382 pgroup = Node.group pnode
1383 sgroup = Node.group snode
1384 in if pgroup /= sgroup
1385 then fail ("Instance placed accross two node groups, primary " ++
1386 show pgroup ++ ", secondary " ++ show sgroup)
1389 -- | Computes the group of an instance per the primary node.
1390 instancePriGroup :: Node.List -> Instance.Instance -> Gdx
1391 instancePriGroup nl i =
1392 let pnode = Container.find (Instance.pNode i) nl
1395 -- | Compute the list of badly allocated instances (split across node
1397 findSplitInstances :: Node.List -> Instance.List -> [Instance.Instance]
1398 findSplitInstances nl =
1399 filter (not . isOk . instanceGroup nl) . Container.elems
1401 -- | Splits a cluster into the component node groups.
1402 splitCluster :: Node.List -> Instance.List ->
1403 [(Gdx, (Node.List, Instance.List))]
1404 splitCluster nl il =
1405 let ngroups = Node.computeGroups (Container.elems nl)
1406 in map (\(guuid, nodes) ->
1407 let nidxs = map Node.idx nodes
1408 nodes' = zip nidxs nodes
1409 instances = Container.filter ((`elem` nidxs) . Instance.pNode) il
1410 in (guuid, (Container.fromList nodes', instances))) ngroups
1412 -- | Compute the list of nodes that are to be evacuated, given a list
1413 -- of instances and an evacuation mode.
1414 nodesToEvacuate :: Instance.List -- ^ The cluster-wide instance list
1415 -> EvacMode -- ^ The evacuation mode we're using
1416 -> [Idx] -- ^ List of instance indices being evacuated
1417 -> IntSet.IntSet -- ^ Set of node indices
1418 nodesToEvacuate il mode =
1419 IntSet.delete Node.noSecondary .
1421 let i = Container.find idx il
1422 pdx = Instance.pNode i
1423 sdx = Instance.sNode i
1424 dt = Instance.diskTemplate i
1425 withSecondary = case dt of
1426 DTDrbd8 -> IntSet.insert sdx ns
1429 ChangePrimary -> IntSet.insert pdx ns
1430 ChangeSecondary -> withSecondary
1431 ChangeAll -> IntSet.insert pdx withSecondary