pump: prevent pump progress from an exit with 0 if the pump become offline
#769
Conversation
|
/run-e2e-in-kind |
weekface
force-pushed
the
fix-pump-offline
branch
from
a8416d5 to
b8d14d9
Compare
|
/run-e2e-in-kind |
| if [ $? == 0 ]; then | ||
| echo $(date -u +"[%Y/%m/%d %H:%M:%S.%3N %:z]") "pump offline, please delete my pod" | ||
| tail -f /dev/null | ||
| fi |
There was a problem hiding this comment.
Can you be sure that if the exit code is 0, it must be pump offline?
There was a problem hiding this comment.
This script ensures that the pump container would not exit normally, it must be killed by kubelet.
There was a problem hiding this comment.
The pause state also ends with exit code 0, but anyway the pod needs to be running.
There was a problem hiding this comment.
What if the container get restarted?
There was a problem hiding this comment.
For example, due to node failure or eviction
There was a problem hiding this comment.
If container restarted, it will be running again, users have to issue binlogctl again to offline the pump. And if the pod is evicted for disk or node failure, then that data may be lost because pump data is not replicated.
|
cherry pick to release-1.0 in PR #778 |
What problem does this PR solve?
prevent pump progress from an exit with 0 if the pump become
offline, because it will be
onlineagain if the pod get restarted bystatefulset.What is changed and how does it work?
Check List
Tests
Code changes
Side effects
Related changes
Does this PR introduce a user-facing change?: