July 23, 20178 yr I have not changed anything on my server but this morning my Dockers (ex. Sonarr, Radarr, and Nextcloud) seem to have lost write access to their containers. I have done a full system reboot but nothing has changed. Fix Common Problems has identified that my cache drive is read only. I have no idea how this might have happened. Any thoughts on how to resolve this? Diagnostics are attached. starkillerbase-diagnostics-20170723-1033.zip
July 23, 20178 yr Tons of errors about corruption on the cache drive. I don't want to give any advice as @johnnie.black is the resident expert / guru / master of all thing BTRFS
July 23, 20178 yr There are a lot of read and write errors on the OCZ cache device, possibly a bad cable, run: btrfs dev stats -z /mnt/cache This will show current error count and reset all values to 0, replace cables to rule them out and then run a correcting scrub, if all errors are corrected monitor for a few days/weeks to make sure all values remain at 0, monitor with: btrfs dev stats /mnt/cache
July 23, 20178 yr Author This seems to have worked for now. I will continue to monitor. Thank you for the insight in troubleshooting btrfs.
July 24, 20178 yr I had a related issue with 6.3.5 - the scary part is not so much that the cache drives file system was set to read-only when a file system error was encountered, but that the unraid GUI did not indicate any error. It really should have a big warning if that happens, at least on the dashboard and array page. I also only noticed after some dockers failed. BTW after fixing the filesystem and rebooting everything was fine again - next time it happens i will try to get diagnostics. Another problem with unraid (at least 6.3.5.) is that if an array data drives has a file system error then the gui in some cases is not startet at all after reboot - the mount spews some error about a corrupt file system in the system log. Here also the unraid webgui should be more resilient and at least start up and give big bright warnings instead of failing to load.
July 24, 20178 yr 23 minutes ago, bonienl said: Do you have notifications enabled? Disk errors are reported thru these. No, but I looked at the gui's pages and none of them indicated an error. Neither the Dashboard nor the Main page. Even with notifications disabled, I would expect a warning or error when looking at the gui. And in case of the corrupt XFS file system the webgui did not even load. Ideally the webgui should alert the user of such fatal occurences and give some guidance in rectifying the problem. A simple warning along the lines "An error occured while mounting the cache drive. It is currently in read only mode. Please see help page for further steps to solve this problem." would have sufficed. And in case of the XFS mount that found a corrupt filesystem it would have been nice if it loaded and produced a similar warning and gave some advice on how to proceed (even if only referring to documentation/forum/etc). Also are you 100% sure systemnotifications catches these? The cache drive was mounted and the array started after all. If not for the dockers barfing, everything would feel normal reading from the array. Edited July 24, 20178 yr by Videodr0me
July 24, 20178 yr I recommend you enable notifications. This system is used to notify the user about anything that requires attention.
July 24, 20178 yr 2 minutes ago, bonienl said: I recommend you enable notifications. This system is used to notify the user about anything that requires attention. This is surely good advice (at the moment system notifications still has the spin-up issue, though). BUT first and foremost the gui should inform the user of such fundamental errors and at least load and be accessable. Systemn hotifications should provide additional safety so one has not to check the gui constantly but I still firmly believe the gui should (1) load in case of mounting errors (2) provide indication that something went wrong and (3) point to dokumentation and/or provide some guidance in fixing the errors. As pretty as the GUI is, I find that it's operation in time of crises is the yardstick to judge it's value especially for storage system such as unraid.
July 24, 20178 yr 1 hour ago, Videodr0me said: This is surely good advice (at the moment system notifications still has the spin-up issue, though). BUT first and foremost the gui should inform the user of such fundamental errors and at least load and be accessable. Systemn hotifications should provide additional safety so one has not to check the gui constantly but I still firmly believe the gui should (1) load in case of mounting errors (2) provide indication that something went wrong and (3) point to dokumentation and/or provide some guidance in fixing the errors. As pretty as the GUI is, I find that it's operation in time of crises is the yardstick to judge it's value especially for storage system such as unraid. While you are correct on pretty much everything here, it is worth noting that the Fix Common Problems plugin would have caught and notified you (once again if notifications are enabled) for much of what you say.
July 24, 20178 yr 41 minutes ago, Squid said: While you are correct on pretty much everything here, it is worth noting that the Fix Common Problems plugin would have caught and notified you (once again if notifications are enabled) for much of what you say. Good catch. I will use that. But still hope this gets fixed so that the gui without any plugins/notifications also reports critical errors. Also are you sure that in case of the corrupt XFS array drive (the case where the gui does not load/work at all) the plugin or notifications still work? Edited July 24, 20178 yr by Videodr0me
July 24, 20178 yr Notifications run independently from the GUI and can be sent to different destinations, like email client or mobile agent. It is not limited to the browser only.
July 24, 20178 yr 2 hours ago, Videodr0me said: sure that in case of the corrupt XFS array drive If its unmountable FCP will notify you. If it gets mounted read-only you'll be notified. If its corrupted enough that the mount command actually crashes and ties up the UI, then probably not. Very difficult situation for me to test.
Archived
This topic is now archived and is closed to further replies.