Closed roedie closed 10 months ago
How many bouncers you got connected and they all using unique api keys?
I've got 28 bouncers connected with unique keys.
The problem mostly seems to occur when the HAProxy instances are reloaded at the same time. When I modify the cronjob to make sure they do not reload at the same time the problem almost goes away.
The problem mostly seems to occur when the HAProxy instances are reloaded at the same time. When I modify the cronjob to make sure they do not reload at the same time the problem almost goes away.
Yeah that wouldn't surprise me as when they load they will ask LAPI for the full blocklist on startup. We have a feature flag on 1.5 that is called chunked_stream
this may help. To enable the feature you can follow the instruction printed when running
cscli config feature-flags
@LaurenceJJones Sorry I didn't respond, I totally missed your reply. First testing seems to show that it helps indeed. Thanks!
@LaurenceJJones Sorry I didn't respond, I totally missed your reply. First testing seems to show that it helps indeed. Thanks!
I'm glad to hear it! If we get enough positive feedback about it, it may become the default.
@LaurenceJJones I just wanted to let you know that my bouncers have been running fine for 2 weeks now with chunked_stream enabled. Thanks again.
@LaurenceJJones I just wanted to let you know that my bouncers have been running fine for 2 weeks now with chunked_stream enabled. Thanks again.
Thank you for the update! This feedback is very helpful as it will push us to make it the default
When running multiple HAProxy bouncers I see this happen randomly:
Not sure if it's a LAPI issue when too much bouncers connect or if it is a bouncer issue. At least a retry would be nice instead of a hard failure.