Konstantin Lebedev

I added additional debug logs, but the reason is still not clear. It is known that restarting the filer helps, but with a start delay of about 5 seconds.

There is just a restart, the filler continues to connect and disconnect in an endless loop `systemctl restart seaweedfs-filer` ``` I0326 09:59:23.964073 filer_grpc_server_sub_meta.go:117 + filer:172.23.1.22:[email protected]:51970 local subscribe / from 2024-03-26...

> the latest version has a fix for this. @chrislusf It would be useful to know in which commit. Perhaps this https://github.com/seaweedfs/seaweedfs/commit/27bb38228b647e34fe20a6016fa04c829138c272 or https://github.com/seaweedfs/seaweedfs/commit/604091a4807a56b7a56caeeb316c0e5a7642e57d ?

I think it will be enough just to add a parameter to the master.toml configuration file to configure the arbitrator.

Add dsn for set [writeTimeout](https://github.com/go-sql-driver/mysql#writetimeout) https://github.com/seaweedfs/seaweedfs/pull/5060

The problem is not reproducible on MySQL 5.7

To check, you can take an already built binary from the branch master https://github.com/seaweedfs/seaweedfs/releases/tag/dev

It seems that you need a flag that will check for the presence of an entry in the filer and skip deleted files, otherwise it will be impossible to re-replicate...

> If there is an error, instead of ignore and advance the timestamp, it should stop and retry after a certain delay? This is easier to implement through a queue...

@chrislusf Plz reopen Now I’ll try to reproduce it and come up with another solution