Lots of files, lots of file actions (ie backup) and lots of free ram and you end up with this.
Doing this every so often (maybe 1x per week) will keep the number reasonable. This was the workaround we did on production nodes that had lots of file activity (MQ creating deleting 500-1000 files/sec) and running for months and typically had 100G of free ram (so the cleanup never had a reason to prune the dentries). echo 2 > /proc/sys/vm/drop_cache Over 50-100million it starts to cause "odd" delays. Ie doing a cat /proc/slabinfo used to pause parts of the kernel for 2-3 seconds when it hits the dentry part. over 500million and I have seen it cause a crash without a umount. And this behavior has been around since at least rhel6 kernels until now (a long time). On Sat, Dec 20, 2025 at 8:27 PM George R Goffe <[email protected]> wrote: > > Roger, > > Thanks for your response... it's APPRECIATED! > > It looks like large dentry values are not surprising given a large storage > device... like one might see with a 24TB filesystem. Yes? > > This was the first and, so far, only time I've seen this message. Even though > I've been doing backups to this device for several months. > > Thanks again for the response. > > Regards, > > George... > > > > > > > > On Friday, December 19, 2025 at 04:14:51 AM PST, Roger Heflin > <[email protected]> wrote: > > > > > > Run this and see if there are any large kernel structures. > > sudo cat /proc/slabinfo | sort -k 15n | tail -10 > > The one I have see a problem with on a filesystem umount (a long > delay) is dentry being greater than 20 million. The higher dentry > is the longer the delay. > > > On Thu, Dec 18, 2025 at 7:03 PM George R Goffe via users > <[email protected]> wrote: > > > > Barry, > > > > From the bugzilla.redhat.com website I got some information which I tried. > > IT WORKED!!! Yay... but, there's more... :( > > > > I ran these commands: > > > > dnf install fedora-repos-rawhide > > > > followed by > > > > dnf update --enablerepo=rawhide kernel > > > > > > These two commands got me to kernel 6.17.12-200.fc42.x86_64+debug > > > > The panic has gone but in its place are some strange error(?) messages that > > are meaningless to me (see below). I couldn't find much info on the web > > about these messages. What I did find is vague suggestions(?) like "update > > your workqueue settings" but just where they are located is STILL a > > mystery. Sigh. > > > > MSGS: > > > > [Thu Dec 18 16:00:32 2025] workqueue: drm_fb_helper_damage_work hogged CPU > > for >10000us 4 times, consider switching to WQ_UNBOUND > > [Thu Dec 18 16:00:32 2025] workqueue: drm_fb_helper_damage_work hogged CPU > > for >10000us 5 times, consider switching to WQ_UNBOUND > > [Thu Dec 18 16:00:33 2025] workqueue: drm_fb_helper_damage_work hogged CPU > > for >10000us 7 times, consider switching to WQ_UNBOUND > > > > > > I got these messages after a umount of a 24TB btrfs filesystem which took a > > (relatively) LONG TIME `1 minute. Usually this happens REALLY quickly. > > Sigh... > > > > A subsequent reboot froze the system (which had trouble finishing shutting > > down) for several minutes. The system finally "woke up" when I powered off > > the 24TB external USB connected hard drive. > > > > I'm trying to submit bugs for abrt which produces a REALLY NICE "details" > > popup... It's NOT "enabled" for copy/paste operations though... Yet another > > bug report. ARGH! > > > > > > > > > -- > > _______________________________________________ > > users mailing list -- [email protected] > > To unsubscribe send an email to [email protected] > > Fedora Code of Conduct: > > https://docs.fedoraproject.org/en-US/project/code-of-conduct/ > > List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines > > List Archives: > > https://lists.fedoraproject.org/archives/list/[email protected] > > Do not reply to spam, report it: > > https://pagure.io/fedora-infrastructure/new_issue > -- _______________________________________________ users mailing list -- [email protected] To unsubscribe send an email to [email protected] Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/[email protected] Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
