[BUG] keydb gets extremely slow when master is not available

[cjabrantes]

Hi all,

Describe the bug Seems to be an issue with keydb when using --active-replica and --replicaof --multi-master. When one instance goes down (and that also makes the dns to not resolve the fqdn) the remaining instance goes extremely slow: time redis-cli -p 6379 ping PONG real 0m28.525s user 0m0.007s sys 0m0.006s

Does anyone have any idea of this situation?

version: keydb:x86_64_v6.0.9

If i set REPLICAOF NO ONE on the remaining instance it start to work fast again.

To reproduce

make that DNS is not aware of the FQDN present of REPLICAOF (in k8s scale down the sts)

Expected behavior

No slowness should be observed just because keydb can't resolve or dont have the master available

Thanks,

[artarik]

can you try use param multi-master-no-forward yes ?

[cjabrantes]

hi @artarik,

Sorry i didn't receive the notification for your updated and as the tickets have so much time i wasn't checking with regularity. I will test and let you know.

Thanks!

[cjabrantes]

hi @artarik, sadly i m not able to reproduce anymore :( with the as is config