Downstream stall notifier fixes #200
Merged
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
On segfault we execute a handler that provides information including a backtrace. This currently emits all information in a single write call after collecting it in a buffer. If anything goes wrong, e.g., the backtrace() call itself crashes, then no information will be emitted. The backtrace() call is not signal safe in theory, and in practice the situation seems mixed as to its safety. So it not unlikely that situations may arise where no output can be emitted on SIGSEGV. Because we catch the signal and then re-raise it using pthread_kill, the specific information about the IP is lost in re-raise: this prevents the line in syslog which usually captures information about segfaults from appearing at all. So we may be left without useful information after a crash. In this change, we emit additional information before the backtrace() which is not likely to have any problem, and we emit each as separate write(2) calls so if there is a failure at any point we at least have the information emitted up to that point. After this, the start of the output on segfault looks like so: Segmentation fault, si_pid: 0, si_addr: 0000000000000000, ip: 0000579ba959751f Segmentation fault resolved ip: 0000000005e2751f in [0000579ba3770000+000000000e3f98d8] Segmentation fault on shard 0, in scheduling group admin. Followed by the backtrace. Closes scylladb#2691 (cherry picked from commit 320f13a)
Use 0x in the addresses in the segfault initial logging to make it clear they are hex (if there happen not to be any letters in the address this can be ambiguous). Closes scylladb#2727 (cherry picked from commit b843f5a)
If a an exception is in progress when a reactor stall is detected, omit taking the backtrace, as this can crash as detailed in scylladb#2697. Add a test which reproduces the issue, and which crashes (sometimes) before this fix and which runs cleanly afterwards. Fixes scylladb#2697. (cherry picked from commit 7f2aede)
StephanDollberg
approved these changes
Apr 15, 2025
This was referenced Apr 15, 2025
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.
This downstreams most of the updates and fixes related to CORE-9596:
Diagnostic upgrades (accepted upstream):
scylladb#2691
scylladb#2727
Crash fix (not accepted upstream yet):
scylladb#2714
I did not downstream one diagnostic fix (accepted upstream) because it does not merge cleanly due to earlier changes to the stall notifier which are not yet in our fork:
scylladb#2712
We will get this fix when we rebase on steastar master.
I plan to backport these to at least 24.2.