Ditto with hardware raid. We had a bug with a SAN firmware update one time. Something to do with "deduplication". Not my area so I don't recall who made the SAN but it was an IBM contract to maintain it. A colleague(*) (much more experienced in such matters) attended all the meetings where it was assured the whole procedure was automatic & painless - and would proceed to do the next SAN and so forth. All manglement wanted to know was how it would affect the business - might be a bit slower so it was scheduled for a weekend when load was lightest.Agree. s*** happens all the time. I'm a big believer in minimum complexity. Although I built the encryption support into sdm (and had fun!), I don't use it here (Pi or otherwise), except for testing. I became a storage simplicity fan when my win2k server MB-based hardware raid disks failed. Completely unrecoverable@bls. I should have omitted luksDump.
Twist of irony considering this thread. My PC failed to boot. Very long story short, systemd noticed the nvme drive efi partition was corrupt (thanks M$) so proceeded to replay the journal and fsck every filesystem. A non-critical filesystem on an ssd failed. Was working fine at reboot. Nothing useful from systemd at the time - just a maintenance shell. Even once the PC was up and running, the cause was unknown. What happened, yes. Cause, no. It does beg the question if systemd on rpi will do the same?
There's a very specific use case for disk encryption. None of them involve the filesystem being mounted.
(*) that friday lunch down the pub he announced "it's no different to a [ahem] virus".
So with a couple of critical people off on holiday, other's out for a night on the town, my colleague took it upon himself that(*) afternoon to wander into the data centre, ?flick a switch, pull a cable, change a config?. Suffice it say the buggy firmware update never made it past the first SAN. Sure everything was in a heap. Headless chickens on saturday morning but the mirror SAN was intact. Stuff was toggled over to that and rebooted. It took days for IBM to come up with a fix.
Software raid wins every time.
Statistics: Posted by swampdog — Thu Jul 11, 2024 2:56 pm