Closed wuase closed 1 month ago
The issue related to the pending pool connection seems to be related to L3 tcp timeout problem discussed in this issue, I'll update and eventually close the issue after some other tests.
@wuase I have the same issuse, is there an analysis conclusion of this issuse
@xingzhang8023 Socket hangup were occurring, kernel fine tuning fixed the issue.
Check out these articles https://blog.cloudflare.com/when-tcp-sockets-refuse-to-die https://access.redhat.com/solutions/726753
Lettuce doesn't seems able to recognize topology changes if the first configured sentinel node goes down with the master. Even configuring a periodic topology refresh seems to be uneffective.
I know from this discussion that looking for master by just one sentinel is a desired behavior, it exists any workaround or procedure to re-establish connection to the correct master?
Update: It seems that the affected connection are the one already established by the pool, new connections works like a charm.
Redis configuration:
Scenario: Spring boot application is connected to Redis by sentinels through lettuce with pool.
Spring boot configuration:
Stacktrace:
Thanks, Angelo