You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
29129: backport-2.1: storage: Up-replicate before removing replicas from dead stores r=a-robinson a=a-robinson
Backport 1/1 commits from #28875.
/cc @cockroachdb/release
---
Fixes #25392, by preventing the following series of events:
1. Node x dies
2. We remove node x's replica of some range
3. Node y dies before we up-replicate, leaving the range unavailable (1
out of 2 replicas dead)
4. Node x comes back online. It can't help the situation, because its
replica was officially removed from the range.
Instead, we now up-replicate first before removing node x's replica.
Release note: None
29177: release-2.1: server: add a separate sampling loop for Go mem stats r=tschottdorf a=petermattis
Backport 5/5 commits from #29061.
/cc @cockroachdb/release
---
`runtime.ReadMemStats` will block waiting for any running GC cycle to
finish. On a large heap this can take upwards of 10s which is in excess of
the default frequency that we sample runtime statistics (10s).
Fixes #27775
Release note: none
Co-authored-by: Alex Robinson <alexdwanerobinson@gmail.com>
Co-authored-by: Peter Mattis <petermattis@gmail.com>