Hoping someone can help because my server is in a state I cannot fix myself.

Background:

spent a couple of days dealing with a flaky SAS expander causing drives to drop. Did some disk swaps to stabilise things, eventually used Tools > New Config to fully remove array slots 14 and 17 (parity-valid ticked). Array itself is fine now - 15 disks DISK_OK, dual parity, completed clean correcting parity check.

The problem:

emhttpd is in a broken state where it actively wipes share state.

All my current shares as missing not showing or mounted, and unable to create any new shares

Symptoms:

- /var/local/emhttp/shares.ini gets written as 0 bytes on every array start

- If I manually populate shares.ini, emhttpd zeros it out again within seconds

- var.ini shows shareCount="0" regardless of how many .cfg files exist

- Shares page in GUI shows nothing

- cmdShareApply and cmdShareDelete via emcmd return "500 Internal Server Error"

- cmdShareAdd does create a .cfg file on flash but doesn't update internal state

- At one point during today's debugging cycles, emhttpd actively DELETED the share .cfg files from /boot/config/shares/ - I have no record of intentionally deleting them via UI

What still works:

- shfs is partially functional - /mnt/user/<sharename>/<path> all work fine for direct access

- ls /mnt/user/ itself returns "Structure needs cleaning" (readdir on root broken)

- All Docker containers using bind mounts to specific share paths work normally

- All data accessible, no data loss

What I've tried:

- Multiple stop/start array cycles - no change

- Multiple full reboots - no change

- Downgraded 7.2.6 to 7.2.5 - identical behaviour on both

- Deleted /boot/config/shares/ and let it regenerate - no change

- Killed shfs and relaunched with -o remember=330 (default seems to be 0) - readdir still broken

- Another New Config - no change

- Manually populated shares.ini from cfg file values - works briefly in UI sort of until emhttpd wipes it, but they didn't mount.

I have a strong suspicion this state was triggered by the New Config that removed slots 14 and 17, but I cannot prove this. I'd not used the shares in a while before that so the broken state could have predated it like when i upgraded to 7.2.6.

Cheers

mediaserver-diagnostics-20260512-1834.zip

Edited May 12May 12 by emub
formatting and correctness.

Syslog in the diags is empty, reboot and post new ones.

Update - resolved.

The empty syslog you flagged was because /boot/config/rsyslog.conf (and /etc/rsyslog.conf) had data corruption - bit-flipped characters throughout config which disabled the file-output rules. Reconstructed a minimal clean config and restarted rsyslogd. I replaced a corrupted USB Stick a long while ago, must have brought over bad config, I don't think syslog has been working for a long time.

The new logs immediately showed the actual culprit: XFS metadata corruption on disk 12, looping every second: XFS (md12p1): Metadata corruption detected.

Stopped array, started in maintenance mode, ran xfs_repair on /dev/md12p1. A few items moved to lost+found, mostly recoverable TV/movie folders. Started array normally and all 11 shares immediately came back.

after xfs_repair on disk 12 and array restart, shares appeared in the GUI but SMB still wasn't exporting them (smb-shares.conf was 0 bytes, shareSMBCount=0). Triggering a single cmdEditShare=Apply via emcmd against any one share caused emhttpd to regenerate smb-shares.conf for all 8 SMB-exported shares, and they're all now visible again.