-
Notifications
You must be signed in to change notification settings - Fork 83
Remove signal handlers #298
Conversation
|
@frerich I think it is safe to do this now but I don't know if it has any side effects when the clcachesrv or memcache is used. |
|
I think I agree that removing the signal handlers should be safe now, but I'll have to think about this some more (and remind myself of the discussions we had back when the handlers were introduced). Thanks for bringing this topic up and providing a patch! |
|
@siu Can you rebase your PR on current master? Thanks! I think that otherwise, this is safe to merge (after dropping Python 3.3 support). |
196d68f to
2714cc0
Compare
|
Done, I have a question though, how does one run the tests locally after the refactor in modules? If I run |
The signal handlers were introduced to avoid cache corruption when the clcache process is stopped in the middle of a write to the cache (statistics, manifests or objects). See frerich#233. Even if SIGINT and SIGTERM were ignored the cache still had a chance to be corrupted in the event of a SIGTERM (which cannot be ignored). Since frerich#233 the writing of files to the cache has been improved to replace the files atomically, reducing the risk of storing corrupted files in the cache. See pull requests frerich#286, frerich#292 and frerich#296. Therefore ignoring these signals is not needed anymore.
2714cc0 to
c9e1e3d
Compare
|
@siu Just run |
|
@xoviat, thank you, I will give it a try |
|
@xoviat that worked. Does that mean that if I forget to run |
|
|
|
Cool, that makes a lot of sense, thanks for clarifying the workflow. |
|
@frerich in my opinion this PR is ready, let me know if anything else is blocking the merge. |
|
Looks good - I suppose this closes #261. |
|
#263 can be closed too. |
The signal handlers were introduced to avoid cache corruption when the
clcache process is stopped in the middle of a write to the cache
(statistics, manifests or objects). See #233. Even if SIGINT and SIGTERM
were ignored the cache still had a chance to be corrupted in the event
of a SIGTERM (which cannot be ignored).
Since #233 the writing of files to the cache has been improved to
replace the files atomically, reducing the risk of storing corrupted
files in the cache. See pull requests #286, #292 and #296. Therefore
ignoring these signals is not needed anymore.