New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
stability: could not GC completed transaction #8978
Comments
These are probably external intent resolutions which reliably end up on the wrong range after splits/rebalances. I thought we fixed that (#7652 (comment)), but perhaps not fully so. |
Ah. I think I know what's going on and it's not good. We construct a GCRequest and give it a key range Take the two keys (the second being the /Local/Range/"\x93"/RangeDescriptor < /Local/Range/"\x93"/"rdsc\x00" Resolving their addresses ( The more general problem here is that while We've talked about making |
Can you expand on what badness this specific problem could cause? Are we just failing to clean up the transaction record for a split? |
We're definitely failing to clean up transaction records for splits and ChangeReplicas (at least for a while; I think the slow GC queue will eventually get them). That may be all that's going wrong, but this has revealed that |
I'll fix the immediate problem. |
The previous way of computing the key range for the GCRequest was incorrectly making the assumption that `(roachpb.Key).Next` respected ordering on the corresponding addressing-resolved keys. This lead to intents on RangeDescriptors not being resolved during range splits and perhaps rebalances, as observed in cockroachdb#8978.
#9036 fixes the immediate problem (and removes the annoying error messages), but I'm still suspicious about this in general. |
This is old and might no longer be an issue. We should verify and close. |
I don't think there's any work remaining here that's not captured in #7880, so I'm closing it. |
Run
zerosum -n 4 -m 3 -c none
(prerequisite,make build). After 30s, kill
zerosum`. The log files will have a mess of the following log message:Immediately before this error, we are performing a split:
I'm not sure what the error means. Perhaps its innocuous, but it is certainly spammy. @tschottdorf?
Cc @cockroachdb/stability
The text was updated successfully, but these errors were encountered: