Question Details

No question body available.

Tags

undo purge polardb

Answers (1)

May 21, 2026 Score: 0 Rep: 1 Quality: Low Completeness: 80%

heyy, dealing with undo log growth can be super annoying haha...

for your questions:

  1. yeah, setting innodbpurgersegtruncatefrequency = 1 will definitely add some noticeable I/O overhead. It forces the system to do truncation checks way too often. I'd bump it up to something like 16 or 32. setting it to 1 is just overkill.

  2. for innodbpurgebatchsize, 1000 is a decent starting point. but if you push it too high (like the 5000 max), yes you'll likely hit some mutex contention and see random cpu spikes. just monitor it.

  3. another parameter you can look at is innodbmaxpurgelag. but be really careful with this one because it will actually slow down your inserts/updates to let the purge catch up. dont use it in prod without testing.

Also just a heads up from experience... no amount of parameter tuning will fix the purge lag if you have super long-running transactions or sleeping connections holding old read views open. double check informationschema.innodbtrx to maake sure there isnt a forgotten query blocking the whole purge process!