Swift processes constantly overload controllers after reboot
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Won't Fix
|
High
|
MOS Swift | ||
Mitaka |
Won't Fix
|
High
|
MOS Swift |
Bug Description
Fuel version info (9.0 mos #427): http://
After all controllers graceful reboot (simultaneous), could becomes broken for 1 hour because different services can't start and work properly due to heavy overload (CPU). After 1 hour my environment was stabilized and OpenStack basic function started to work (OSTF passed), but all controllers are still overloaded by Swift python processes:
http://
node specs: http://
Also Swift processes are being constantly re-spawned by something (PIDs are changed):
http://
Here is a part of swift logs:
http://
Steps to reproduce:
1. Create cluster with 3 controllers and 2 computes, defaults for storage
2. Reboot all cluster nodes
3. Run OSTF
Expected result: tests are passed
Actual result: test fails for ~ 1 hour due to controller nodes overload
Diagnostic snapshot: https:/
Changed in fuel: | |
assignee: | Fuel Sustaining (fuel-sustaining-team) → MOS Swift (mos-swift) |
Changed in fuel: | |
status: | New → Confirmed |
tags: | added: swarm-fail |
Changed in fuel: | |
status: | Confirmed → Won't Fix |
tags: | added: move-to-9.2 |
As for me, it is to risky to try to fix it quickly before HCF, because:
- tested case is not usual (rebooting of all nodes), it means, that issue is reproduced only in specific scenario, not after every deployment;
- finally, env goes to "working" status (> "After 1 hour my environment was stabilized and OpenStack basic function started to work (OSTF passed)")