Filer is not working properly with cassandra cluster

[t4nmoy]

I have 3 nodes Cassandra cluster and it's used by a filer server. I have configured filer for super large directories. If I take one of the Cassandra nodes down then sometimes read request on filer does not find the requested file until I restart the filer. After restarting the filer it works normally with the two Cassandra nodes and can find the file.

[chrislusf]

SeaweedFS version?

[t4nmoy]

SeaweedFS version: 2.60

[chrislusf]

Added a fix 9d85569 for this.

[t4nmoy]

Thanks

[nivekuil]

If restarting the filer makes a difference then I think the problem here is probably not cassandra consistency but the client not reconnecting, as in gocql/gocql#1555 or gocql/gocql#915

[chrislusf]

@nivekuil Thanks for checking and your analysis is very reasonable.

Seems we need to upgrade the gocql version?