IP set full error in kernel log
Bug #1439817 reported by
Salvatore Orlando
This bug affects 3 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Fix Released
|
High
|
Brian Haley | ||
Juno |
Fix Released
|
Undecided
|
Unassigned |
Bug Description
This is appearing in some logs upstream: http://
And it has also been reported by andreaf in IRC as having been observed downstream.
Logstash is not very helpful as this manifests only with a job currently in the experimental queue.
As said job runs in non-isolated mode, accruing of elements in the IPset until it reaches saturation is onet things that might need to be investigated.
Changed in neutron: | |
importance: | Undecided → High |
Changed in neutron: | |
milestone: | none → kilo-rc1 |
Changed in neutron: | |
status: | Fix Committed → Fix Released |
Changed in neutron: | |
milestone: | kilo-rc1 → 2015.1.0 |
To post a comment you must log in.
The default number of elements in an ipset set is only 64K, so it can be hit pretty easily during some test cycles (we've seen it at HP now obviously). As a quick fix we can make both the hashelem and hashsize values tunable, so that a larger size can be set. A better long-term solution might be to detect this failure and increase the size automatically.
I'll assign this to myself and get a quick patch out for discussion.