Fix #10080 - OSD crashes during msgr reconnection #3029
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Let us say OSD A and B are peers with each other, at certain point, B is marked down by monitor (and thus kicked off osd map), but the daemon is still there, A would mark the connection with B down and clean everything. When B is in again, it will try to connect with A and negotiate the message in/out sequence. At A side, it will receive B's in_seq and drop messages whose sequence is equal to or less than B's in_seq. In this case, A's out_seq might be inconsistent to B's in_seq. Solution: At A side, after receiving B's in_seq, if after locally clean-up, if its out_seq is still less than B's in_seq, bump up its out_seq to make two sides consistent with each other. Signed-off-by: Guang Yang <yguang@yahoo-inc.com>
ghost
added
|
It seemed reasonable for the situation. But the solution maybe better that A reply a reset tag to let B clean up its state and issue connect msg again by checking connect_seq or in_seq? |
|
It seems to depend on the underlying assumption in such case:
My understanding is that our case belongs to 1 (please correct me if my I am wrong here). One step forward might to reset B's in_seq to 0, but it is a little bit harder (from implementation's perspective). |
|
There's something else wrong here. I'll discuss it more on the tracker, but for the brief reason: if we lost some session state, then this is a flimsy band-aid and I think that no matter the underlying cause it will be missing things. :( |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Event flow analysis:
Let us say OSD A and B are peers with each other, at certain point, B is marked down by monitor (and thus kicked off osd map), but the daemon is still there, A would mark the connection with B down and clean everything. When B is in again, it will try to connect with A and negotiate the message in/out sequence. At A side, it will receive B's in_seq and drop messages whose sequence is equal to or less than B's in_seq. In this case, A's out_seq might be inconsistent to B's in_seq.
Solution:
At A side, after receiving B's in_seq, if after locally clean-up, if its out_seq is still less than B's in_seq, bump up its out_seq to make two sides consistent with each other.
Fixes: #10080
Signed-off-by: Guang Yang yguang@yahoo-inc.com