pt-archiver throttling to prevent Galera node Lag
General
Escalation
General
Escalation
Description
Environment
Galera XtraDB cluster by Percona on Ubuntu, three physical nodes via a 20Gbps fiber
Activity
Show:
Aaditya Dubey December 25, 2024 at 3:23 PM
Hi @Sander Lootens
Thank you for the report.
Closing the report as you found the solution to the described issue.
Sander Lootens December 22, 2024 at 10:47 AM
Someone pointed me to the max-flow-ctl-parameter, which should do exactly what my problem describes. So my request is already accomplished.
Done
Details
Details
Assignee
Unassigned
UnassignedReporter
Sander Lootens
Sander LootensPriority
LowAffects versions
Labels
Smart Checklist
Open Smart Checklist
Smart Checklist
Open Smart Checklist
Created December 20, 2024 at 10:32 AM
Updated December 25, 2024 at 3:24 PM
Resolved December 25, 2024 at 3:23 PM
As a DBA, I would like
pt-archiverto automatically throttle its operations when synchronization lag is detected in Galera Cluster nodes to maintain optimal performance and data consistency during archiving tasks.We recently experienced a 15-second lag between Galera nodes caused by pt-archiver deleting 10,000 records per commit. Although the transaction size was within
wsrep_max_ws_size, the operation caused issues with a table containing large records (having aMEDIUMTEXT). Existing parameters like--sleep-coefand--max-lagdid not help, and other monitoring tools failed to detect the problem since overall performance appeared stable.