Quesnelia - Error (No space left on device) during CALL delete_old_marc_indexers_versions()
Description
CSP Request Details
CSP approval requested on 2/5/2025
CSP Rejection Details
None
Potential Workaround
None
blocks
clones
Checklist
hideActivity
Show:
Done
Details
Details
Assignee

Reporter

Priority
Story Points
0
Sprint
None
Development Team
Folijet
Fix versions
Release
Quesnelia (R1 2024) Service Patch #10
RCA Group
Implementation coding issue
CSP Approved
Yes
TestRail: Cases
Open TestRail: Cases
TestRail: Runs
Open TestRail: Runs
Created February 3, 2025 at 7:28 PM
Updated February 7, 2025 at 3:12 PM
Resolved February 4, 2025 at 10:14 AM
TestRail: Cases
TestRail: Runs
With update of mod-source-record-storage PTF observed regular high spikes of CPU usage on DB. After investigation we found that it’s CALL delete_old_marc_indexers_versions() that’s happening on regular basis (by default it’s 30 min).
In logs (in module logs and DB logs) errors found each time when this DB job is trying to clean fs09 tenant (ptf primary tenant) : io.vertx.pgclient.PgException: ERROR: could not write to file "base/pgsql_tmp/pgsql_tmp5862.422": No space left on device (53100).
As discovered that may happen when data set to be cleaned by DB query may is too big.
Version: mod-source-record-storage:5.9.7
Data Set size: PTF data set marc.indexers 2.18 B records
To avoid hanging the system at large data volumes, the possibility to limit the number of old records deleted simultaneously has been added:
Added the possibility of periodic startup and scheduled startup to reduce the amount of CPU time consumed: