History | View | Annotate | Download (54.1 kB)
Show list of pending acquires in “gnt-debug locks”
This is accomplished by keeping a list of waiting threads insteadof just their number inside the lock-internal condition. A fewother tweaks to the output format are also made.
Signed-off-by: Michael Hanselmann <hansmi@google.com>...
Fix race condition in locking unittest
While writing unittests for the new lock monitor, I made a small mistake anddidn't synchronize acquiring locks properly. This patch takes care of thisissue by adding additional synchronization.
Add simple lock monitor
This patch adds an initial implementation of a lock monitor, accessiblefor the user through “gnt-debug locks”. It currently shows all resourcelocks: BGL, nodes and instances. Config and job queue locks could beshown too, but wouldn't be of much help. The current owner(s) and mode...
Implement lock names for debugging purposes
This patch adds lock names to SharedLocks and LockSets, that can be usedlater for displaying the actual locks being held/used in places where weonly have the lock, and not the entire context of the locking operation....
User assertFalse instead of assert_(not ...)
Signed-off-by: Manuel Franceschini <livewire@google.com>Reviewed-by: Guido Trotter <ultrotter@google.com>
_BaseCondition: allow saving/restoring state
Signed-off-by: Guido Trotter <ultrotter@google.com>Reviewed-by: Michael Hanselmann <hansmi@google.com>
SharedLock _acquire_restore and _release_save
If a shared lock is used inside a condition, we need to make sure thatit's reacquired in the same way as it was originally, after the wait.
Fix long-standing race condition bug in locking unittest
Every once in a while we saw failures in the locking unittest, butweren't able to determine the cause. This patch fixes it by usingthreading.Event instead of plain threading.Condition to notify another...
Ignore log messages in unittests
Signed-off-by: Michael Hanselmann <hansmi@google.com>Reviewed-by: Iustin Pop <iustin@google.com>
Test LockSet.acquire return value for timeout
Signed-off-by: Michael Hanselmann <hansmi@google.com>Reviewed-by: Guido Trotter <ultrotter@google.com>
locking.LockSet: Implement acquire timeouts
The timeout passed to LockSet.acquire() is measured over all lock acquires. IfLockSet.acquire fails to acquire all requested locks within the specifiedamount of time, all locks are released again and the acquire fails....
Replace all xrange() with range()
More locking tests race conditions fixes
There were more race conditions. By adding a notify function toSharedLock.acquire we can prevent them.
SingleActionPipeCondition =~ s/Action/Notify/
With this patch we simplify usage on the SingleActionCondition (whichwasn't a condition at all) by making it a real condition. This way wecan just wait() on it, or notifyAll() as we would on a normal one. The...
testNotification: add more checking about order
Abstract base condition test cases
This way they can be used to test different condition classes.
Move the "done" queue inside _ThreadedTestCase
All (ok, all but one) _ThreadedTestCase users have a done Queue, so wemove its building in the _ThreadedTestCase setUp
locking.SharedLock: Fix bug in delete function
SharedLock.__acquire_unlocked uses keyword parameters. Just passingthe timeout would set the “shared” parameter.
Rename LockSet.acquire parameter “blocking” to “timeout”
Also remove the “blocking” parameter from LockSet.remove andGanetiLockManager.remove. There's no point in implementing timeouts on removalunless we need them.
Try to fix locking unittests
Our automated test system found a few problems in the new lockingunittests. This patch should fix them, although I wasn't able toreproduce the problem. All are race conditions.
Add _PipeCondition class
_PipeCondition is a condition implemented using pipe(2) and poll(2).It allows the implementation of timeouts without using a busy-wait loopwith time.sleep.
Unlike Python's built-in threading.Condition class and to save filedescriptors and an internal queue, it can only be used to notify...
Add _SingleActionPipeCondition class
This class will be used as a basic block for pipe(2)-basedconditions. Upon initialization it creates a pipe and can benotified once (hence the “single action” in the name). Acallable helper class is used to wait for notifications....
SharedLock: implement timeouts
This patch greatly simplifies the SharedLock code and implementstimeouts for the acquire() and delete() functions. A wrapper aroundPython's threading.Condition class must be used to ensure threadsafety when check whether there are any waiters left....
Skip another testWaitingXblocksY
The waiting sharer blocks exclusive is another not-possible to testright now.
Sorry for missing it the first time; there are no other testWaiting...tests.
Reviewed-by: ultrotter
Skip the waitingExclusiveBlockSharer test
Until we have a way to properly test this, we'll skip the test.
Reviewed-by: imsnah
Convert the locking unittests to repetition-test
Currently the locking tests are using timeouts to ensure an event will'never happen'. However, this is suboptimal.
The patch converts all of these to instead: not wait, but sequence theoperations logically and expect that they execute as such. In case of...
LockSet: forbid add() on a partially owned set
This patch bans add() on a half-acquired set. This behavior waspreviously possible, but created a deadlock if someone tried to acquirethe set-lock in the meantime, and thus is now forbidden. ThetestAddRemove unit test is fixed for this new behavior, and includes a...
Fix LockSet._names() to work with the set-lock
If the set-lock is acquired, currently, the _names function will fail ona double acquire of a non-recursive lock. This patch fixes the behavior,and some lines of code added to the testAcquireSetLock test check that...
A few more locking unit tests
A few more tests written while bug-hunting. One of them shows a realissue, at last. :)
Add lock-all-through-GLM unit test
I was hunting for a bug in my code and thought the culprit was in thelocking library, so I added a test to check. Unfortunately turns out itwasn't. :( Committing the test anyway, while still trying to figure outwhat's wrong......
Invert nodes/instances locking order
An implementation mistake from the original design caused nodes to belocked before instances, rather than after. This patch inverts the levelnumbering, changing also the relevant unittests and the recursivelocking function starting point....
Add a more comment lines to testLockingConstants
This is to discourage even more whoever may think that this requirementis not really useful and can be lifted, and to at least know where it'sused before trying to break it.
Add a new LockSet unittest
This test checks the LockSet behaviour when an empty list is passed.The current behaviour is expected, but since this is a corner case,we're safer to keep it under a check, and if we need a different onemonitor that everything is as we expect it to be....
Locking: remove LEVEL_CONFIG lockset
Since the ConfigWriter now handles its own locking it's not necessary tohave a specific level for the config in the Locking Manager anymore.This patch thus removes it, and all the unittest calls that used it, ordepended on it being present....
Locking: add ssynchronized decorator
This patch creates a new decorator function ssynchronized in the lockinglibrary, which takes as input a SharedLock, and synchronizes access tothe decorated functions using it. The usual SharedLock semantics apply,so it's possible to call more than one synchronized function at the same...
locking: remove obsolete comment
Reviewed-by: iustinp
LockSet: handle empty case
A LockSet is mostly useful when it has some locks in it. On the other handthere are cases in which it must function even when empty. For example if acluster has no instances in it there's no reason why locking all of themshouldn't work anyway. This patch adds test code for that situation and...
LockSet: add missing check code
This check that no operation had been performed before release() was missing inthe test code. Adding it.
Codestyle updates for locking code
LockSet: make acquire() able to get the whole set
This new functionality makes it possible to acquire a whole set, by passing"None" to the acquire() function as the list of elements. This will avoid newadditions to the set, and then acquire all the current elements. The list of...
LockSet: improve remove() api
Lockset's remove() function used to return a list of locks we failed to remove.Rather than doing this we'll return a list of removed locks, so it's moresimilar to how acquire() behaves. This patch also fixes the relevant unit tests....
LockSet: make acquire() return the set of names
In a LockSet acquire() returned True on success. This code changes that toreturn a set containing the names of the elements acquired. This is still atrue value if we acquired any lock but is slightly more useful (because if...
Initial GanetiLockManager implementation
Includes some locking-related constants and explanations on how theLockManager should be used, the class itself and its test cases.
The class includes: - a basic constructor - functions to acquire and release lists of locks at the same level...
LockSet implementation and unit tests
A LockSet represents locking for a set of resources of the same type. A threadcan acquire multiple resources at the same time, and release some or all ofthem, but cannot acquire more resources incrementally at different times...
Add a few SharedLock delete() tests
- Check that even a shared acquire() fails on a deleted lock- Check that delete() fails on a lock you share (must own it or nothing)
These are assumptions I build on in future code, so better check for them.Currently no code change is necessary for them to be valid....
SharedLock: fix a wrong unit-test helper code
The _doItDelete helper code was supposed to be used to dispatch threads thatdeleted the SharedLock. It actually just acquired it exclusively. This remainedunnoticed as the helper thread is just used to test interaction, not the delete...
Add the delete() operation to SharedLock
This new operation lets a lock be cleanly deleted. The lock will be exclusivelyheld before deletion, and after it pending and future acquires will raise anexception. Other SharedLock operations are modify to deal with delete() and to...
Fix a couple of SharedLock docstrings
Use the actual class name rather than a spaced version of it.
Shared Lock implementation and unit tests.
Adding a locking.py file for the ganeti locking library. Its first component isthe implementation of a non-recursive blocking shared lock complete with atesting library.
Reviewed-by: imsnah, iustinp