issue-1751: [Filestore] Handle errors received by WriteData requests in Flush #4793
Conversation
e673
force-pushed
the
users/nasonov/issue-1751-handle-errors
branch
from
18c468c to
76ee2bf
Compare
|
Note This is an automated comment that will be appended during run. 🔴 linux-x86_64-relwithdebinfo: some tests FAILED for commit 76ee2bf.
🔴 linux-x86_64-relwithdebinfo: some tests FAILED for commit 76ee2bf.
🔴 linux-x86_64-relwithdebinfo: some tests FAILED for commit 76ee2bf.
|
e673
force-pushed
the
users/nasonov/issue-1751-handle-errors
branch
from
76ee2bf to
a98bb37
Compare
e673
force-pushed
the
users/nasonov/issue-1751-handle-errors
branch
from
a98bb37 to
099ce7d
Compare
| } | ||
|
|
||
| if (WriteBackCache) { | ||
| // TODO |
There was a problem hiding this comment.
Q: What should I do if FlushNodeData returned non-retriable error?
There was a problem hiding this comment.
I would say that fsync(fd) and sync() should return errors until the problematic fd is closed
| Stats->FlushCompleted(); | ||
| CompleteFlush(nodeState); | ||
| if (nonRetriableError) { | ||
| // TODO(#1751): handle non-retriable errors |
There was a problem hiding this comment.
Q: What should I do if it is not possible to flush?
There was a problem hiding this comment.
keep problematic fd's data in the cache until this fd is closed, then just discard its data
There was a problem hiding this comment.
you can also experiment with ext4 and see what it does when it fails to flush (e.g. in case the fs is out of space)
#1751
Current behavior:
Expected behavior:
Discussion needed: