I've checked for bad blocks on all disks that are part of the RAID array connected to my SAS controller: none found, which in itself is a good thing, but still doesn't provide an explanation about the server freezes. As it's a RAID 5 array, I could only take out one disk, check it for bad blocks, re-add it to the array, wait for the disk to have re-synced with the array, before I could go ahead with the next disk. That all takes quite some time.
There's also the possibility to check for bad blocks with a read/write test, instead of just read-only. This test takes a whole lot longer to complete, but should I have done it anyway? What's the advantage of this test over the read-only test? Does it find bad blocks that would otherwise not be found?
I'm running with the Samba daemon stopped for a few weeks now. No freezes anymore, but making do without Samba is a real pain, as this means there's no NAS on the LAN anymore. That's why checking for bad blocks took over two weeks to complete, because I was copying a folder of more than 2 TiBs over SFTP to my desktop pc, which makes the re-sync on the RAID array very, very slow. As it does seem to be the culprit, but I still refuse to believe the freezes are caused by software but instead by hardware, there are only two possibilities left in my opinion: the SAS controller or the CPU, with the SAS controller being the most likely candidate. Sigh, decisions, decisions…
