Scheduled Maintenance: We are aware of an issue with Google, AOL, and Yahoo services as email providers which are blocking new registrations. We are trying to fix the issue and we have several internal and external support tickets in process to resolve the issue. Please see: viewtopic.php?t=158230

 

 

 

[Software] Serious deadlock issue jbd2

Linux Kernel, Network, and Services configuration.
Message
Author
User avatar
Penaut Butter
Posts: 13
Joined: 2018-05-31 21:59
Has thanked: 1 time

Re: [Software] Serious deadlock issue jbd2

#41 Post by Penaut Butter »

Just a quick update from my side, just for the record.
I ended up fixing it accidentally. See, my HDD got this SMR technology (i was not familiar with it) which seems to exhibit poor reads and writes after some time of heavy I/O (IME, having decent speeds for some time, and after around 8 to 20GB of writes the speed would drop to a single digit value[2mbps]).
I got some info from here and there, being https://wiki.archlinux.org/title/Advanced_Format the most informative one.

I did the following:
Originally I had luks+xfs, I went ahead and reformatted the xfs part with "mkfs.xfs -s size=4096 /dev/device" but the issue was still there, so checked my luks dump and noticed it had a sector size of 512 (which causes SMRs to be severely and negatively impacted), I tried to "cryptsetup reencrypt --sector-size=4096 device" but the process was so slow I decided to simply, wipe the disk properly, and then to "cryptsetup luksFormat --sector-size=4096 device" and after that I tried both xfs and ext4 with 4096 sector size and the bug was GONE for both of them (i wanted to be sure more than 1 fs was working correctly)
(i decided to use ext4 from now on as a personal choice, it seems to have improved a lot since the last time i used it years ago, great!).

My conclusion would be that if one has a SMR HDD (bad luck), the best thing to do is to set 'sector size' to 4096 to whatever the FS you're going to format to (even luks if you're encrypting), and to use the latest kernel/format tools to create that new FS. Because my drive not only had the wrong sector size values but it was formatted with an old versions of luks and xfs (keeping in mind that both linux kernel and some filesystems added patches to deal with SMR devices the last few years).

Aki
Global Moderator
Global Moderator
Posts: 2979
Joined: 2014-07-20 18:12
Location: Europe
Has thanked: 75 times
Been thanked: 407 times

Re: [Software] Serious deadlock issue jbd2

#42 Post by Aki »

Hello @Penaut Butter,
Penaut Butter wrote: 2024-02-26 21:46 [..] my HDD got this SMR technology (i was not familiar with it) which seems to exhibit poor reads and writes after some time of heavy I/O (IME, having decent speeds for some time, and after around 8 to 20GB of writes the speed would drop to a single digit value[2mbps]). I got some info from here and there, being https://wiki.archlinux.org/title/Advanced_Format the most informative one. [..] My conclusion would be that if one has a SMR HDD (bad luck), the best thing to do is to set 'sector size' to 4096 to whatever the FS you're going to format to [..] (keeping in mind that both linux kernel and some filesystems added patches to deal with SMR devices the last few years).
Therefore, in your case, the deadlock was caused by a hardware problem (hard drive with SMR technology) related to the selected sector size.

I would be nice to get an updated from the original poster @rontarplan if his issue was completely solved.
⢀⣴⠾⠻⢶⣦⠀
⣾⠁⢠⠒⠀⣿⡁ Debian - The universal operating system
⢿⡄⠘⠷⠚⠋⠀ https://www.debian.org
⠈⠳⣄⠀

Post Reply