root / src / Ganeti / JQueue.hs @ ae858516
History | View | Annotate | Download (13.8 kB)
1 |
{-# LANGUAGE TemplateHaskell #-} |
---|---|
2 |
|
3 |
{-| Implementation of the job queue. |
4 |
|
5 |
-} |
6 |
|
7 |
{- |
8 |
|
9 |
Copyright (C) 2010, 2012 Google Inc. |
10 |
|
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. |
15 |
|
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. |
20 |
|
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 |
24 |
02110-1301, USA. |
25 |
|
26 |
-} |
27 |
|
28 |
module Ganeti.JQueue |
29 |
( QueuedOpCode(..) |
30 |
, QueuedJob(..) |
31 |
, InputOpCode(..) |
32 |
, queuedOpCodeFromMetaOpCode |
33 |
, queuedJobFromOpCodes |
34 |
, Timestamp |
35 |
, noTimestamp |
36 |
, opStatusFinalized |
37 |
, extractOpSummary |
38 |
, calcJobStatus |
39 |
, calcJobPriority |
40 |
, jobFileName |
41 |
, liveJobFile |
42 |
, archivedJobFile |
43 |
, determineJobDirectories |
44 |
, getJobIDs |
45 |
, sortJobIDs |
46 |
, loadJobFromDisk |
47 |
, noSuchJob |
48 |
, readSerialFromDisk |
49 |
, allocateJobIds |
50 |
, allocateJobId |
51 |
) where |
52 |
|
53 |
import Control.Concurrent.MVar |
54 |
import Control.Exception |
55 |
import Control.Monad |
56 |
import Data.List |
57 |
import Data.Maybe |
58 |
import Data.Ord (comparing) |
59 |
-- workaround what seems to be a bug in ghc 7.4's TH shadowing code |
60 |
import Prelude hiding (log, id) |
61 |
import System.Directory |
62 |
import System.FilePath |
63 |
import System.IO.Error (isDoesNotExistError) |
64 |
import System.Posix.Files |
65 |
import qualified Text.JSON |
66 |
import Text.JSON.Types |
67 |
|
68 |
import Ganeti.BasicTypes |
69 |
import qualified Ganeti.Constants as C |
70 |
import Ganeti.JSON |
71 |
import Ganeti.Logging |
72 |
import Ganeti.Objects (Node) |
73 |
import Ganeti.OpCodes |
74 |
import Ganeti.Path |
75 |
import Ganeti.Rpc (executeRpcCall, RpcCallJobqueueUpdate(..)) |
76 |
import Ganeti.THH |
77 |
import Ganeti.Types |
78 |
import Ganeti.Utils |
79 |
|
80 |
-- * Data types |
81 |
|
82 |
-- | The ganeti queue timestamp type |
83 |
type Timestamp = (Int, Int) |
84 |
|
85 |
-- | Missing timestamp type. |
86 |
noTimestamp :: Timestamp |
87 |
noTimestamp = (-1, -1) |
88 |
|
89 |
-- | An input opcode. |
90 |
data InputOpCode = ValidOpCode MetaOpCode -- ^ OpCode was parsed successfully |
91 |
| InvalidOpCode JSValue -- ^ Invalid opcode |
92 |
deriving (Show, Eq) |
93 |
|
94 |
-- | JSON instance for 'InputOpCode', trying to parse it and if |
95 |
-- failing, keeping the original JSValue. |
96 |
instance Text.JSON.JSON InputOpCode where |
97 |
showJSON (ValidOpCode mo) = Text.JSON.showJSON mo |
98 |
showJSON (InvalidOpCode inv) = inv |
99 |
readJSON v = case Text.JSON.readJSON v of |
100 |
Text.JSON.Error _ -> return $ InvalidOpCode v |
101 |
Text.JSON.Ok mo -> return $ ValidOpCode mo |
102 |
|
103 |
-- | Invalid opcode summary. |
104 |
invalidOp :: String |
105 |
invalidOp = "INVALID_OP" |
106 |
|
107 |
-- | Tries to extract the opcode summary from an 'InputOpCode'. This |
108 |
-- duplicates some functionality from the 'opSummary' function in |
109 |
-- "Ganeti.OpCodes". |
110 |
extractOpSummary :: InputOpCode -> String |
111 |
extractOpSummary (ValidOpCode metaop) = opSummary $ metaOpCode metaop |
112 |
extractOpSummary (InvalidOpCode (JSObject o)) = |
113 |
case fromObjWithDefault (fromJSObject o) "OP_ID" ("OP_" ++ invalidOp) of |
114 |
Just s -> drop 3 s -- drop the OP_ prefix |
115 |
Nothing -> invalidOp |
116 |
extractOpSummary _ = invalidOp |
117 |
|
118 |
$(buildObject "QueuedOpCode" "qo" |
119 |
[ simpleField "input" [t| InputOpCode |] |
120 |
, simpleField "status" [t| OpStatus |] |
121 |
, simpleField "result" [t| JSValue |] |
122 |
, defaultField [| [] |] $ |
123 |
simpleField "log" [t| [(Int, Timestamp, ELogType, JSValue)] |] |
124 |
, simpleField "priority" [t| Int |] |
125 |
, optionalNullSerField $ |
126 |
simpleField "start_timestamp" [t| Timestamp |] |
127 |
, optionalNullSerField $ |
128 |
simpleField "exec_timestamp" [t| Timestamp |] |
129 |
, optionalNullSerField $ |
130 |
simpleField "end_timestamp" [t| Timestamp |] |
131 |
]) |
132 |
|
133 |
$(buildObject "QueuedJob" "qj" |
134 |
[ simpleField "id" [t| JobId |] |
135 |
, simpleField "ops" [t| [QueuedOpCode] |] |
136 |
, optionalNullSerField $ |
137 |
simpleField "received_timestamp" [t| Timestamp |] |
138 |
, optionalNullSerField $ |
139 |
simpleField "start_timestamp" [t| Timestamp |] |
140 |
, optionalNullSerField $ |
141 |
simpleField "end_timestamp" [t| Timestamp |] |
142 |
]) |
143 |
|
144 |
-- | Convenience function to obtain a QueuedOpCode from a MetaOpCode |
145 |
queuedOpCodeFromMetaOpCode :: MetaOpCode -> QueuedOpCode |
146 |
queuedOpCodeFromMetaOpCode op = |
147 |
QueuedOpCode { qoInput = ValidOpCode op |
148 |
, qoStatus = OP_STATUS_QUEUED |
149 |
, qoPriority = opSubmitPriorityToRaw . opPriority . metaParams |
150 |
$ op |
151 |
, qoLog = [] |
152 |
, qoResult = JSNull |
153 |
, qoStartTimestamp = Nothing |
154 |
, qoEndTimestamp = Nothing |
155 |
, qoExecTimestamp = Nothing |
156 |
} |
157 |
|
158 |
-- | From a job-id and a list of op-codes create a job. This is |
159 |
-- the pure part of job creation, as allocating a new job id |
160 |
-- lives in IO. |
161 |
queuedJobFromOpCodes :: (Monad m) => JobId -> [MetaOpCode] -> m QueuedJob |
162 |
queuedJobFromOpCodes jobid ops = do |
163 |
ops' <- mapM (`resolveDependencies` jobid) ops |
164 |
return QueuedJob { qjId = jobid |
165 |
, qjOps = map queuedOpCodeFromMetaOpCode ops' |
166 |
, qjReceivedTimestamp = Nothing |
167 |
, qjStartTimestamp = Nothing |
168 |
, qjEndTimestamp = Nothing |
169 |
} |
170 |
|
171 |
-- | Job file prefix. |
172 |
jobFilePrefix :: String |
173 |
jobFilePrefix = "job-" |
174 |
|
175 |
-- | Computes the filename for a given job ID. |
176 |
jobFileName :: JobId -> FilePath |
177 |
jobFileName jid = jobFilePrefix ++ show (fromJobId jid) |
178 |
|
179 |
-- | Parses a job ID from a file name. |
180 |
parseJobFileId :: (Monad m) => FilePath -> m JobId |
181 |
parseJobFileId path = |
182 |
case stripPrefix jobFilePrefix path of |
183 |
Nothing -> fail $ "Job file '" ++ path ++ |
184 |
"' doesn't have the correct prefix" |
185 |
Just suffix -> makeJobIdS suffix |
186 |
|
187 |
-- | Computes the full path to a live job. |
188 |
liveJobFile :: FilePath -> JobId -> FilePath |
189 |
liveJobFile rootdir jid = rootdir </> jobFileName jid |
190 |
|
191 |
-- | Computes the full path to an archives job. BROKEN. |
192 |
archivedJobFile :: FilePath -> JobId -> FilePath |
193 |
archivedJobFile rootdir jid = |
194 |
let subdir = show (fromJobId jid `div` C.jstoreJobsPerArchiveDirectory) |
195 |
in rootdir </> jobQueueArchiveSubDir </> subdir </> jobFileName jid |
196 |
|
197 |
-- | Map from opcode status to job status. |
198 |
opStatusToJob :: OpStatus -> JobStatus |
199 |
opStatusToJob OP_STATUS_QUEUED = JOB_STATUS_QUEUED |
200 |
opStatusToJob OP_STATUS_WAITING = JOB_STATUS_WAITING |
201 |
opStatusToJob OP_STATUS_SUCCESS = JOB_STATUS_SUCCESS |
202 |
opStatusToJob OP_STATUS_RUNNING = JOB_STATUS_RUNNING |
203 |
opStatusToJob OP_STATUS_CANCELING = JOB_STATUS_CANCELING |
204 |
opStatusToJob OP_STATUS_CANCELED = JOB_STATUS_CANCELED |
205 |
opStatusToJob OP_STATUS_ERROR = JOB_STATUS_ERROR |
206 |
|
207 |
-- | Computes a queued job's status. |
208 |
calcJobStatus :: QueuedJob -> JobStatus |
209 |
calcJobStatus QueuedJob { qjOps = ops } = |
210 |
extractOpSt (map qoStatus ops) JOB_STATUS_QUEUED True |
211 |
where |
212 |
terminalStatus OP_STATUS_ERROR = True |
213 |
terminalStatus OP_STATUS_CANCELING = True |
214 |
terminalStatus OP_STATUS_CANCELED = True |
215 |
terminalStatus _ = False |
216 |
softStatus OP_STATUS_SUCCESS = True |
217 |
softStatus OP_STATUS_QUEUED = True |
218 |
softStatus _ = False |
219 |
extractOpSt [] _ True = JOB_STATUS_SUCCESS |
220 |
extractOpSt [] d False = d |
221 |
extractOpSt (x:xs) d old_all |
222 |
| terminalStatus x = opStatusToJob x -- abort recursion |
223 |
| softStatus x = extractOpSt xs d new_all -- continue unchanged |
224 |
| otherwise = extractOpSt xs (opStatusToJob x) new_all |
225 |
where new_all = x == OP_STATUS_SUCCESS && old_all |
226 |
|
227 |
-- | Determine whether an opcode status is finalized. |
228 |
opStatusFinalized :: OpStatus -> Bool |
229 |
opStatusFinalized = (> OP_STATUS_RUNNING) |
230 |
|
231 |
-- | Compute a job's priority. |
232 |
calcJobPriority :: QueuedJob -> Int |
233 |
calcJobPriority QueuedJob { qjOps = ops } = |
234 |
helper . map qoPriority $ filter (not . opStatusFinalized . qoStatus) ops |
235 |
where helper [] = C.opPrioDefault |
236 |
helper ps = minimum ps |
237 |
|
238 |
-- | Log but ignore an 'IOError'. |
239 |
ignoreIOError :: a -> Bool -> String -> IOError -> IO a |
240 |
ignoreIOError a ignore_noent msg e = do |
241 |
unless (isDoesNotExistError e && ignore_noent) . |
242 |
logWarning $ msg ++ ": " ++ show e |
243 |
return a |
244 |
|
245 |
-- | Compute the list of existing archive directories. Note that I/O |
246 |
-- exceptions are swallowed and ignored. |
247 |
allArchiveDirs :: FilePath -> IO [FilePath] |
248 |
allArchiveDirs rootdir = do |
249 |
let adir = rootdir </> jobQueueArchiveSubDir |
250 |
contents <- getDirectoryContents adir `Control.Exception.catch` |
251 |
ignoreIOError [] False |
252 |
("Failed to list queue directory " ++ adir) |
253 |
let fpaths = map (adir </>) $ filter (not . ("." `isPrefixOf`)) contents |
254 |
filterM (\path -> |
255 |
liftM isDirectory (getFileStatus (adir </> path)) |
256 |
`Control.Exception.catch` |
257 |
ignoreIOError False True |
258 |
("Failed to stat archive path " ++ path)) fpaths |
259 |
|
260 |
-- | Build list of directories containing job files. Note: compared to |
261 |
-- the Python version, this doesn't ignore a potential lost+found |
262 |
-- file. |
263 |
determineJobDirectories :: FilePath -> Bool -> IO [FilePath] |
264 |
determineJobDirectories rootdir archived = do |
265 |
other <- if archived |
266 |
then allArchiveDirs rootdir |
267 |
else return [] |
268 |
return $ rootdir:other |
269 |
|
270 |
-- Function equivalent to the \'sequence\' function, that cannot be used because |
271 |
-- of library version conflict on Lucid. |
272 |
-- FIXME: delete this and just use \'sequence\' instead when Lucid compatibility |
273 |
-- will not be required anymore. |
274 |
sequencer :: [Either IOError [JobId]] -> Either IOError [[JobId]] |
275 |
sequencer l = fmap reverse $ foldl seqFolder (Right []) l |
276 |
|
277 |
-- | Folding function for joining multiple [JobIds] into one list. |
278 |
seqFolder :: Either IOError [[JobId]] |
279 |
-> Either IOError [JobId] |
280 |
-> Either IOError [[JobId]] |
281 |
seqFolder (Left e) _ = Left e |
282 |
seqFolder (Right _) (Left e) = Left e |
283 |
seqFolder (Right l) (Right el) = Right $ el:l |
284 |
|
285 |
-- | Computes the list of all jobs in the given directories. |
286 |
getJobIDs :: [FilePath] -> IO (Either IOError [JobId]) |
287 |
getJobIDs paths = liftM (fmap concat . sequencer) (mapM getDirJobIDs paths) |
288 |
|
289 |
-- | Sorts the a list of job IDs. |
290 |
sortJobIDs :: [JobId] -> [JobId] |
291 |
sortJobIDs = sortBy (comparing fromJobId) |
292 |
|
293 |
-- | Computes the list of jobs in a given directory. |
294 |
getDirJobIDs :: FilePath -> IO (Either IOError [JobId]) |
295 |
getDirJobIDs path = do |
296 |
either_contents <- |
297 |
try (getDirectoryContents path) :: IO (Either IOError [FilePath]) |
298 |
case either_contents of |
299 |
Left e -> do |
300 |
logWarning $ "Failed to list job directory " ++ path ++ ": " ++ show e |
301 |
return $ Left e |
302 |
Right contents -> do |
303 |
let jids = foldl (\ids file -> |
304 |
case parseJobFileId file of |
305 |
Nothing -> ids |
306 |
Just new_id -> new_id:ids) [] contents |
307 |
return . Right $ reverse jids |
308 |
|
309 |
-- | Reads the job data from disk. |
310 |
readJobDataFromDisk :: FilePath -> Bool -> JobId -> IO (Maybe (String, Bool)) |
311 |
readJobDataFromDisk rootdir archived jid = do |
312 |
let live_path = liveJobFile rootdir jid |
313 |
archived_path = archivedJobFile rootdir jid |
314 |
all_paths = if archived |
315 |
then [(live_path, False), (archived_path, True)] |
316 |
else [(live_path, False)] |
317 |
foldM (\state (path, isarchived) -> |
318 |
liftM (\r -> Just (r, isarchived)) (readFile path) |
319 |
`Control.Exception.catch` |
320 |
ignoreIOError state True |
321 |
("Failed to read job file " ++ path)) Nothing all_paths |
322 |
|
323 |
-- | Failed to load job error. |
324 |
noSuchJob :: Result (QueuedJob, Bool) |
325 |
noSuchJob = Bad "Can't load job file" |
326 |
|
327 |
-- | Loads a job from disk. |
328 |
loadJobFromDisk :: FilePath -> Bool -> JobId -> IO (Result (QueuedJob, Bool)) |
329 |
loadJobFromDisk rootdir archived jid = do |
330 |
raw <- readJobDataFromDisk rootdir archived jid |
331 |
-- note: we need some stricness below, otherwise the wrapping in a |
332 |
-- Result will create too much lazyness, and not close the file |
333 |
-- descriptors for the individual jobs |
334 |
return $! case raw of |
335 |
Nothing -> noSuchJob |
336 |
Just (str, arch) -> |
337 |
liftM (\qj -> (qj, arch)) . |
338 |
fromJResult "Parsing job file" $ Text.JSON.decode str |
339 |
|
340 |
-- | Read the job serial number from disk. |
341 |
readSerialFromDisk :: IO (Result JobId) |
342 |
readSerialFromDisk = do |
343 |
filename <- jobQueueSerialFile |
344 |
tryAndLogIOError (readFile filename) "Failed to read serial file" |
345 |
(makeJobIdS . rStripSpace) |
346 |
|
347 |
-- | Allocate new job ids. |
348 |
-- To avoid races while accessing the serial file, the threads synchronize |
349 |
-- over a lock, as usual provided by an MVar. |
350 |
allocateJobIds :: [Node] -> MVar () -> Int -> IO (Result [JobId]) |
351 |
allocateJobIds mastercandidates lock n = |
352 |
if n <= 0 |
353 |
then return . Bad $ "Can only allocate positive number of job ids" |
354 |
else do |
355 |
takeMVar lock |
356 |
rjobid <- readSerialFromDisk |
357 |
case rjobid of |
358 |
Bad s -> do |
359 |
putMVar lock () |
360 |
return . Bad $ s |
361 |
Ok jid -> do |
362 |
let current = fromJobId jid |
363 |
serial_content = show (current + n) ++ "\n" |
364 |
serial <- jobQueueSerialFile |
365 |
write_result <- try $ atomicWriteFile serial serial_content |
366 |
:: IO (Either IOError ()) |
367 |
putMVar lock () |
368 |
case write_result of |
369 |
Left e -> do |
370 |
let msg = "Failed to write serial file: " ++ show e |
371 |
logError msg |
372 |
return . Bad $ msg |
373 |
Right () -> do |
374 |
_ <- executeRpcCall mastercandidates |
375 |
$ RpcCallJobqueueUpdate serial serial_content |
376 |
return $ mapM makeJobId [(current+1)..(current+n)] |
377 |
|
378 |
-- | Allocate one new job id. |
379 |
allocateJobId :: [Node] -> MVar () -> IO (Result JobId) |
380 |
allocateJobId mastercandidates lock = do |
381 |
jids <- allocateJobIds mastercandidates lock 1 |
382 |
return (jids >>= monadicThe "Failed to allocate precisely one Job ID") |