MrBurns977
Members
-
Joined
-
Last visited
-
Since around Unraid 6.10 (exact version unclear, but confirmed gone by the 6.10 release candidates), the internal kernel field rdevLastIO that the Dynamix Cache Directories plugin relies on has been removed. This field was previously used to determine the last disk I/O activity. As a result, the plugin can no longer properly detect when array disks have been accessed. This leads to two kinds of problems: Errors and warnings in the system log when the plugin tries to read rdevLastIO. Incorrect behavior, because disk activity is not recognized reliably. This can trigger unnecessary spin-ups, or conversely, the plugin may believe disks are idle when they are not. I have prepared a fix that replaces the outdated rdevLastIO check with a new method based on /sys/block/*/stat. The plugin now maintains a lightweight state file to track per-device counters and timestamps, which restores correct activity detection on current Unraid versions. A pull request with this fix is available here: https://github.com/unraid/cache_dirs/pull/1
-
I've made a few adjustments to the fnc_time_since_last_disk_access() function (see attachment) to address issues with rdevLastIO no longer being reliable in recent Unraid versions. The updated version uses /sys/block/sd* to track disk I/O activity more accurately and handles more than 26 sdX devices. I've tested it on my setup, and it appears to be working reliably now. What would be the best way to submit this fix upstream for inclusion in future versions? SNIPPET-fnc_time_since_last_disk_access-new.sh
-
I am running the latest Dynamix Cache Directories plugin (version 2025.03.06) on Unraid 7.1.4 and was trying to play with the settings and enabled logging. I soon noticed that my disks were always shown as Idle and idle before/after scan was constant displayed as 9998s/9998s in /var/log/cache_dirs.log. Digging into the code of cache_dirs script and the function fnc_time_since_last_disk_access which is supposed to get the seconds elapsed after the last disk access, I saw that this function tries to find rdevLastIO in the output of /usr/local/sbin/mdcmd status. This variable / statistic is not there anymore and so this function does not work and always returns 9998. See here https://github.com/unraid/cache_dirs/blob/master/source/scripts/cache_dirs#L581 Is this a known problem? Is there any other statistic we can use to fix this problem?