Ok. This is not final solution. But.
- If you checked permissions on tmp folder and found it right then problem with high iowait (iostat usefull utility to diag this correlating error message with read/write metrics, also try to do read/write benchmarks on your backup storage).
- Any solution that you can apply to reduce iowait will be helpful in your situation, but for me I only achieved after moved to ZFS (At this point monitoring system no more alerting me on high iowait but high cpu load warnings still exist). In latest kernels btrfs got many fixes that I didn’t tested, you can try google something like “btrfs optimization”.
- There is my urbackup + ubuntu 20.04 server setup. This is for Hyper-V VM, all options based on my experience, be carefully and adapt for your environment before usage if you wishes. I still working on it.
What I can’t resolve: shortly still found no way to manage memory, cache, buffer for urbackup server.
Of course it’s working fine with small count of clients. But if we talking about tens then issues occurs.
In /backupdata/urbackup_tmp_files directory filled with this cps.* files. When advanced options to store it on system default tmp chosen, then system drive fully filled as result system crashes.Now I planning to configure urbackup user quota and test this settings. Better of course to have something like MAX_BUFFER_SIZE in urbackup server options but I didn’t found this option.
- But putting buffer storage on different location (faster than same backup storage) that I think can be done very simply can improve backup jobs performance. And I wondered why this is not configurable.