Details
-
Bug
-
Resolution: Fixed
-
Major
-
None
-
Untriaged
-
0
-
Unknown
Description
As seen in the #ns_server slack: https://couchbase.slack.com/archives/CLAEKRV33/p1724078831030499
There was a question about behavior of cbcollect_info.. it was quickly decided that it was a bug and eventually I came across it. The main problem is how LiteralTask's are getting ran:
Old Cluster:
==============================================================================
|
Couchbase config
|
|
==============================================================================
|
[{alert_limits,
|
[{max_overhead_perc,50},{max_disk_used,90},{max_indexer_ram,75}]},
|
{audit,
|
[{auditd_enabled,false},
|
{rotate_interval,86400},
|
This cluster:
==============================================================================
|
Couchbase config
|
[{alert_limits,
|
[{max_overhead_perc,50},{max_disk_used,90},{max_indexer_ram,75}]},
|
{audit,
|
[{auditd_enabled,false},
|
{rotate_interval,86400},
|
So it appears that the "literal" that is meant to be logged only to the output is being interpolated into the header as well. You can see this by scrolling to the end of the couchbase config and realizing that it's printed again (like normal).
The issue appears to be an erroneous call to super()._init_() in a class that shouldn't have done that. That change came in through: https://review.couchbase.org/c/ns_server/+/213760
Attachments
Issue Links
- is caused by
-
MB-62714 cbcollect should not attempt to redact prometheus data
- Resolved