We use EOS version 5.1.30 with guarkdb and qrain replicas.
We encountered problems with file corruption on our EOS cluster. We found this out when we checked the checksums of the recorded files with the original files. Presumably the problem occurs when one of the replicas changes. In this regard, we have the following questions.
- When files are corrupted, the following messages appear in the logs: msg=“no local fst metadata present”. What do they mean?
- Can a broken replica corrupt a file? What happens when one of the replicas goes bad?
- Is it possible to view the contents of the replicas (checks with checksums)?
- Is it possible to restore files using the correct replicas?