1 {-| Module describing a node.
3 All updates are functional (copy-based) and return a new node with
9 Copyright (C) 2009, 2010, 2011 Google Inc.
11 This program is free software; you can redistribute it and/or modify
12 it under the terms of the GNU General Public License as published by
13 the Free Software Foundation; either version 2 of the License, or
14 (at your option) any later version.
16 This program is distributed in the hope that it will be useful, but
17 WITHOUT ANY WARRANTY; without even the implied warranty of
18 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 General Public License for more details.
21 You should have received a copy of the GNU General Public License
22 along with this program; if not, write to the Free Software
23 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
28 module Ganeti.HTools.Node
33 -- ** Finalization after data loading
48 -- * Instance (re)location
61 , conflictingPrimaries
74 import Data.List hiding (group)
75 import qualified Data.Map as Map
76 import qualified Data.Foldable as Foldable
77 import Data.Ord (comparing)
78 import Text.Printf (printf)
80 import qualified Ganeti.HTools.Container as Container
81 import qualified Ganeti.HTools.Instance as Instance
82 import qualified Ganeti.HTools.PeerMap as P
84 import qualified Ganeti.HTools.Types as T
86 -- * Type declarations
88 -- | The tag map type.
89 type TagMap = Map.Map String Int
93 { name :: String -- ^ The node name
94 , alias :: String -- ^ The shortened name (for display purposes)
95 , tMem :: Double -- ^ Total memory (MiB)
96 , nMem :: Int -- ^ Node memory (MiB)
97 , fMem :: Int -- ^ Free memory (MiB)
98 , xMem :: Int -- ^ Unaccounted memory (MiB)
99 , tDsk :: Double -- ^ Total disk space (MiB)
100 , fDsk :: Int -- ^ Free disk space (MiB)
101 , tCpu :: Double -- ^ Total CPU count
102 , uCpu :: Int -- ^ Used VCPU count
103 , pList :: [T.Idx] -- ^ List of primary instance indices
104 , sList :: [T.Idx] -- ^ List of secondary instance indices
105 , idx :: T.Ndx -- ^ Internal index for book-keeping
106 , peers :: P.PeerMap -- ^ Pnode to instance mapping
107 , failN1 :: Bool -- ^ Whether the node has failed n1
108 , rMem :: Int -- ^ Maximum memory needed for failover by
109 -- primaries of this node
110 , pMem :: Double -- ^ Percent of free memory
111 , pDsk :: Double -- ^ Percent of free disk
112 , pRem :: Double -- ^ Percent of reserved memory
113 , pCpu :: Double -- ^ Ratio of virtual to physical CPUs
114 , mDsk :: Double -- ^ Minimum free disk ratio
115 , mCpu :: Double -- ^ Max ratio of virt-to-phys CPUs
116 , loDsk :: Int -- ^ Autocomputed from mDsk low disk
118 , hiCpu :: Int -- ^ Autocomputed from mCpu high cpu
120 , offline :: Bool -- ^ Whether the node should not be used
121 -- for allocations and skipped from score
123 , utilPool :: T.DynUtil -- ^ Total utilisation capacity
124 , utilLoad :: T.DynUtil -- ^ Sum of instance utilisation
125 , pTags :: TagMap -- ^ Map of primary instance tags and their count
126 , group :: T.Gdx -- ^ The node's group (index)
127 } deriving (Show, Read, Eq)
129 instance T.Element Node where
134 allNames n = [name n, alias n]
136 -- | A simple name for the int, node association list.
137 type AssocList = [(T.Ndx, Node)]
139 -- | A simple name for a node map.
140 type List = Container.Container Node
142 -- | A simple name for an allocation element (here just for logistic
144 type AllocElement = (List, Instance.Instance, [Node], T.Score)
146 -- | Constant node index for a non-moveable instance.
150 -- * Helper functions
152 -- | Add a tag to a tagmap.
153 addTag :: TagMap -> String -> TagMap
154 addTag t s = Map.insertWith (+) s 1 t
156 -- | Add multiple tags.
157 addTags :: TagMap -> [String] -> TagMap
158 addTags = foldl' addTag
160 -- | Adjust or delete a tag from a tagmap.
161 delTag :: TagMap -> String -> TagMap
162 delTag t s = Map.update (\v -> if v > 1
167 -- | Remove multiple tags.
168 delTags :: TagMap -> [String] -> TagMap
169 delTags = foldl' delTag
171 -- | Check if we can add a list of tags to a tagmap.
172 rejectAddTags :: TagMap -> [String] -> Bool
173 rejectAddTags t = any (`Map.member` t)
175 -- | Check how many primary instances have conflicting tags. The
176 -- algorithm to compute this is to sum the count of all tags, then
177 -- subtract the size of the tag map (since each tag has at least one,
178 -- non-conflicting instance); this is equivalent to summing the
179 -- values in the tag map minus one.
180 conflictingPrimaries :: Node -> Int
181 conflictingPrimaries (Node { pTags = t }) = Foldable.sum t - Map.size t
183 -- * Initialization functions
185 -- | Create a new node.
187 -- The index and the peers maps are empty, and will be need to be
188 -- update later via the 'setIdx' and 'buildPeers' functions.
189 create :: String -> Double -> Int -> Int -> Double
190 -> Int -> Double -> Bool -> T.Gdx -> Node
191 create name_init mem_t_init mem_n_init mem_f_init
192 dsk_t_init dsk_f_init cpu_t_init offline_init group_init =
193 Node { name = name_init
208 , pMem = fromIntegral mem_f_init / mem_t_init
209 , pDsk = fromIntegral dsk_f_init / dsk_t_init
212 , offline = offline_init
214 , mDsk = T.defReservedDiskRatio
215 , mCpu = T.defVcpuRatio
216 , loDsk = mDskToloDsk T.defReservedDiskRatio dsk_t_init
217 , hiCpu = mCpuTohiCpu T.defVcpuRatio cpu_t_init
218 , utilPool = T.baseUtil
219 , utilLoad = T.zeroUtil
224 -- | Conversion formula from mDsk\/tDsk to loDsk.
225 mDskToloDsk :: Double -> Double -> Int
226 mDskToloDsk mval tdsk = floor (mval * tdsk)
228 -- | Conversion formula from mCpu\/tCpu to hiCpu.
229 mCpuTohiCpu :: Double -> Double -> Int
230 mCpuTohiCpu mval tcpu = floor (mval * tcpu)
232 -- | Changes the index.
234 -- This is used only during the building of the data structures.
235 setIdx :: Node -> T.Ndx -> Node
236 setIdx t i = t {idx = i}
238 -- | Changes the alias.
240 -- This is used only during the building of the data structures.
241 setAlias :: Node -> String -> Node
242 setAlias t s = t { alias = s }
244 -- | Sets the offline attribute.
245 setOffline :: Node -> Bool -> Node
246 setOffline t val = t { offline = val }
248 -- | Sets the unnaccounted memory.
249 setXmem :: Node -> Int -> Node
250 setXmem t val = t { xMem = val }
252 -- | Sets the max disk usage ratio.
253 setMdsk :: Node -> Double -> Node
254 setMdsk t val = t { mDsk = val, loDsk = mDskToloDsk val (tDsk t) }
256 -- | Sets the max cpu usage ratio.
257 setMcpu :: Node -> Double -> Node
258 setMcpu t val = t { mCpu = val, hiCpu = mCpuTohiCpu val (tCpu t) }
260 -- | Computes the maximum reserved memory for peers from a peer map.
261 computeMaxRes :: P.PeerMap -> P.Elem
262 computeMaxRes = P.maxElem
264 -- | Builds the peer map for a given node.
265 buildPeers :: Node -> Instance.List -> Node
268 (\i_idx -> let inst = Container.find i_idx il
269 mem = if Instance.autoBalance inst
270 then Instance.mem inst
272 in (Instance.pNode inst, mem))
274 pmap = P.accumArray (+) mdata
275 new_rmem = computeMaxRes pmap
276 new_failN1 = fMem t <= new_rmem
277 new_prem = fromIntegral new_rmem / tMem t
278 in t {peers=pmap, failN1 = new_failN1, rMem = new_rmem, pRem = new_prem}
280 -- | Assigns an instance to a node as primary and update the used VCPU
281 -- count, utilisation data and tags map.
282 setPri :: Node -> Instance.Instance -> Node
283 setPri t inst = t { pList = Instance.idx inst:pList t
285 , pCpu = fromIntegral new_count / tCpu t
286 , utilLoad = utilLoad t `T.addUtil` Instance.util inst
287 , pTags = addTags (pTags t) (Instance.tags inst)
289 where new_count = uCpu t + Instance.vcpus inst
291 -- | Assigns an instance to a node as secondary without other updates.
292 setSec :: Node -> Instance.Instance -> Node
293 setSec t inst = t { sList = Instance.idx inst:sList t
294 , utilLoad = old_load { T.dskWeight = T.dskWeight old_load +
295 T.dskWeight (Instance.util inst) }
297 where old_load = utilLoad t
299 -- * Update functions
301 -- | Sets the free memory.
302 setFmem :: Node -> Int -> Node
304 let new_n1 = new_mem <= rMem t
305 new_mp = fromIntegral new_mem / tMem t
306 in t { fMem = new_mem, failN1 = new_n1, pMem = new_mp }
308 -- | Removes a primary instance.
309 removePri :: Node -> Instance.Instance -> Node
311 let iname = Instance.idx inst
312 new_plist = delete iname (pList t)
313 new_mem = fMem t + Instance.mem inst
314 new_dsk = fDsk t + Instance.dsk inst
315 new_mp = fromIntegral new_mem / tMem t
316 new_dp = fromIntegral new_dsk / tDsk t
317 new_failn1 = new_mem <= rMem t
318 new_ucpu = uCpu t - Instance.vcpus inst
319 new_rcpu = fromIntegral new_ucpu / tCpu t
320 new_load = utilLoad t `T.subUtil` Instance.util inst
321 in t { pList = new_plist, fMem = new_mem, fDsk = new_dsk
322 , failN1 = new_failn1, pMem = new_mp, pDsk = new_dp
323 , uCpu = new_ucpu, pCpu = new_rcpu, utilLoad = new_load
324 , pTags = delTags (pTags t) (Instance.tags inst) }
326 -- | Removes a secondary instance.
327 removeSec :: Node -> Instance.Instance -> Node
329 let iname = Instance.idx inst
330 pnode = Instance.pNode inst
331 new_slist = delete iname (sList t)
332 new_dsk = fDsk t + Instance.dsk inst
334 old_peem = P.find pnode old_peers
335 new_peem = if Instance.autoBalance inst
336 then old_peem - Instance.mem inst
338 new_peers = if new_peem > 0
339 then P.add pnode new_peem old_peers
340 else P.remove pnode old_peers
342 new_rmem = if old_peem < old_rmem
344 else computeMaxRes new_peers
345 new_prem = fromIntegral new_rmem / tMem t
346 new_failn1 = fMem t <= new_rmem
347 new_dp = fromIntegral new_dsk / tDsk t
348 old_load = utilLoad t
349 new_load = old_load { T.dskWeight = T.dskWeight old_load -
350 T.dskWeight (Instance.util inst) }
351 in t { sList = new_slist, fDsk = new_dsk, peers = new_peers
352 , failN1 = new_failn1, rMem = new_rmem, pDsk = new_dp
353 , pRem = new_prem, utilLoad = new_load }
355 -- | Adds a primary instance (basic version).
356 addPri :: Node -> Instance.Instance -> T.OpResult Node
357 addPri = addPriEx False
359 -- | Adds a primary instance (extended version).
360 addPriEx :: Bool -- ^ Whether to override the N+1 and
361 -- other /soft/ checks, useful if we
362 -- come from a worse status
364 -> Node -- ^ The target node
365 -> Instance.Instance -- ^ The instance to add
366 -> T.OpResult Node -- ^ The result of the operation,
367 -- either the new version of the node
369 addPriEx force t inst =
370 let iname = Instance.idx inst
371 new_mem = fMem t - Instance.mem inst
372 new_dsk = fDsk t - Instance.dsk inst
373 new_failn1 = new_mem <= rMem t
374 new_ucpu = uCpu t + Instance.vcpus inst
375 new_pcpu = fromIntegral new_ucpu / tCpu t
376 new_dp = fromIntegral new_dsk / tDsk t
378 new_load = utilLoad t `T.addUtil` Instance.util inst
379 inst_tags = Instance.tags inst
383 _ | new_mem <= 0 -> T.OpFail T.FailMem
384 | new_dsk <= 0 -> T.OpFail T.FailDisk
385 | mDsk t > new_dp && strict -> T.OpFail T.FailDisk
386 | new_failn1 && not (failN1 t) && strict -> T.OpFail T.FailMem
387 | l_cpu >= 0 && l_cpu < new_pcpu && strict -> T.OpFail T.FailCPU
388 | rejectAddTags old_tags inst_tags -> T.OpFail T.FailTags
390 let new_plist = iname:pList t
391 new_mp = fromIntegral new_mem / tMem t
392 r = t { pList = new_plist, fMem = new_mem, fDsk = new_dsk
393 , failN1 = new_failn1, pMem = new_mp, pDsk = new_dp
394 , uCpu = new_ucpu, pCpu = new_pcpu
395 , utilLoad = new_load
396 , pTags = addTags old_tags inst_tags }
399 -- | Adds a secondary instance (basic version).
400 addSec :: Node -> Instance.Instance -> T.Ndx -> T.OpResult Node
401 addSec = addSecEx False
403 -- | Adds a secondary instance (extended version).
404 addSecEx :: Bool -> Node -> Instance.Instance -> T.Ndx -> T.OpResult Node
405 addSecEx force t inst pdx =
406 let iname = Instance.idx inst
409 new_dsk = fDsk t - Instance.dsk inst
410 secondary_needed_mem = if Instance.autoBalance inst
411 then Instance.mem inst
413 new_peem = P.find pdx old_peers + secondary_needed_mem
414 new_peers = P.add pdx new_peem old_peers
415 new_rmem = max (rMem t) new_peem
416 new_prem = fromIntegral new_rmem / tMem t
417 new_failn1 = old_mem <= new_rmem
418 new_dp = fromIntegral new_dsk / tDsk t
419 old_load = utilLoad t
420 new_load = old_load { T.dskWeight = T.dskWeight old_load +
421 T.dskWeight (Instance.util inst) }
424 _ | new_dsk <= 0 -> T.OpFail T.FailDisk
425 | mDsk t > new_dp && strict -> T.OpFail T.FailDisk
426 | secondary_needed_mem >= old_mem && strict -> T.OpFail T.FailMem
427 | new_failn1 && not (failN1 t) && strict -> T.OpFail T.FailMem
429 let new_slist = iname:sList t
430 r = t { sList = new_slist, fDsk = new_dsk
431 , peers = new_peers, failN1 = new_failn1
432 , rMem = new_rmem, pDsk = new_dp
433 , pRem = new_prem, utilLoad = new_load }
438 -- | Computes the amount of available disk on a given node.
439 availDisk :: Node -> Int
447 -- | Computes the amount of used disk on a given node.
449 iDsk t = truncate (tDsk t) - fDsk t
451 -- | Computes the amount of available memory on a given node.
452 availMem :: Node -> Int
460 -- | Computes the amount of available memory on a given node.
461 availCpu :: Node -> Int
469 -- | The memory used by instances on a given node.
471 iMem t = truncate (tMem t) - nMem t - xMem t - fMem t
473 -- * Display functions
475 -- | Return a field for a given node.
476 showField :: Node -- ^ Node which we're querying
477 -> String -- ^ Field name
478 -> String -- ^ Field value as string
481 "idx" -> printf "%4d" $ idx t
484 "status" -> case () of
488 "tmem" -> printf "%5.0f" $ tMem t
489 "nmem" -> printf "%5d" $ nMem t
490 "xmem" -> printf "%5d" $ xMem t
491 "fmem" -> printf "%5d" $ fMem t
492 "imem" -> printf "%5d" $ iMem t
493 "rmem" -> printf "%5d" $ rMem t
494 "amem" -> printf "%5d" $ fMem t - rMem t
495 "tdsk" -> printf "%5.0f" $ tDsk t / 1024
496 "fdsk" -> printf "%5d" $ fDsk t `div` 1024
497 "tcpu" -> printf "%4.0f" $ tCpu t
498 "ucpu" -> printf "%4d" $ uCpu t
499 "pcnt" -> printf "%3d" $ length (pList t)
500 "scnt" -> printf "%3d" $ length (sList t)
501 "plist" -> show $ pList t
502 "slist" -> show $ sList t
503 "pfmem" -> printf "%6.4f" $ pMem t
504 "pfdsk" -> printf "%6.4f" $ pDsk t
505 "rcpu" -> printf "%5.2f" $ pCpu t
506 "cload" -> printf "%5.3f" uC
507 "mload" -> printf "%5.3f" uM
508 "dload" -> printf "%5.3f" uD
509 "nload" -> printf "%5.3f" uN
510 "ptags" -> intercalate "," . map (uncurry (printf "%s=%d")) .
512 "peermap" -> show $ peers t
515 T.DynUtil { T.cpuWeight = uC, T.memWeight = uM,
516 T.dskWeight = uD, T.netWeight = uN } = utilLoad t
518 -- | Returns the header and numeric propery of a field.
519 showHeader :: String -> (String, Bool)
522 "idx" -> ("Index", True)
523 "name" -> ("Name", False)
524 "fqdn" -> ("Name", False)
525 "status" -> ("F", False)
526 "tmem" -> ("t_mem", True)
527 "nmem" -> ("n_mem", True)
528 "xmem" -> ("x_mem", True)
529 "fmem" -> ("f_mem", True)
530 "imem" -> ("i_mem", True)
531 "rmem" -> ("r_mem", True)
532 "amem" -> ("a_mem", True)
533 "tdsk" -> ("t_dsk", True)
534 "fdsk" -> ("f_dsk", True)
535 "tcpu" -> ("pcpu", True)
536 "ucpu" -> ("vcpu", True)
537 "pcnt" -> ("pcnt", True)
538 "scnt" -> ("scnt", True)
539 "plist" -> ("primaries", True)
540 "slist" -> ("secondaries", True)
541 "pfmem" -> ("p_fmem", True)
542 "pfdsk" -> ("p_fdsk", True)
543 "rcpu" -> ("r_cpu", True)
544 "cload" -> ("lCpu", True)
545 "mload" -> ("lMem", True)
546 "dload" -> ("lDsk", True)
547 "nload" -> ("lNet", True)
548 "ptags" -> ("PrimaryTags", False)
549 "peermap" -> ("PeerMap", False)
550 -- TODO: add node fields (group.uuid, group)
551 _ -> (T.unknownField, False)
553 -- | String converter for the node list functionality.
554 list :: [String] -> Node -> [String]
555 list fields t = map (showField t) fields
558 -- | Constant holding the fields we're displaying by default.
559 defaultFields :: [String]
561 [ "status", "name", "tmem", "nmem", "imem", "xmem", "fmem"
562 , "rmem", "tdsk", "fdsk", "tcpu", "ucpu", "pcnt", "scnt"
563 , "pfmem", "pfdsk", "rcpu"
564 , "cload", "mload", "dload", "nload" ]
566 -- | Split a list of nodes into a list of (node group UUID, list of
567 -- associated nodes).
568 computeGroups :: [Node] -> [(T.Gdx, [Node])]
569 computeGroups nodes =
570 let nodes' = sortBy (comparing group) nodes
571 nodes'' = groupBy (\a b -> group a == group b) nodes'
572 in map (\nl -> (group (head nl), nl)) nodes''