[SOLVED] Re: Riddling activity on encrypted and mounted partition
Hi, Andy Smith wrote: > your reply doesn't make it clear to me whether the > lazy init was the cause of your writes or not. It seems so. The disk is mounted without i/o being counted in /sys/block/sda/sda2/stat . If only half of the 733702 write ops of mkfs.ext4 were due to lazy_*=0, then they would keep with lazy_*=1 the disk busy for more than 6 hours at 16 ops per second. This matches my initial observation that it lasted for at least one hour. But in the end i can only say for sure that the drive LED behaves like before i mounted the lazily created ext4 in its encrypted partition. I had ext4 before. So my suspicions were rather directed towards the encryption, which is new for me. Now the whole story matches lazy ext4 creation, regardless of encryption. Have a nice day :) Thomas
Re: [SOLVED] Re: Riddling activity on encrypted and mounted partition
On Tue, Sep 29, 2020 at 01:02:35PM +0200, Thomas Schmitt wrote: > Andy Smith wrote: > > Create with: > >mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 > > This lasts significantly longer than my first mkfs run. > The drive makes ~ 1950 write operations per second. So i estimate that > the job would have lasted hours with ~ 16 writes per second. > In the end mkfs.ext4 caused 733702 write ops on the 3.6 TB partition. > > Ok. New UUID into fstab ... mount ... mkdir ... touch ... Yay ! > > The i/o is still lazy (no wonder with 32 GB RAM), but after about a minute > i see no newly counted writes. > > Thanks a lot ! No problem, but your reply doesn't make it clear to me whether the lazy init was the cause of your writes or not. Maybe I just lack the reading comprehension. Cheers, Andy -- https://bitfolk.com/ -- No-nonsense VPS hosting
[SOLVED] Re: Riddling activity on encrypted and mounted partition
Hi, Andy Smith wrote: > Could it possibly be the lazy init feature of ext4, which is enabled > by default and can sometimes result in several minutes of background > writes to a newly-created fs? Well, the blinking went on for at least an hour. > Create with: >mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 This lasts significantly longer than my first mkfs run. The drive makes ~ 1950 write operations per second. So i estimate that the job would have lasted hours with ~ 16 writes per second. In the end mkfs.ext4 caused 733702 write ops on the 3.6 TB partition. Ok. New UUID into fstab ... mount ... mkdir ... touch ... Yay ! The i/o is still lazy (no wonder with 32 GB RAM), but after about a minute i see no newly counted writes. Thanks a lot ! Have a nice day :) Thomas
Re: Riddling activity on encrypted and mounted partition
Hello, On Tue, Sep 29, 2020 at 10:24:44AM +0200, Thomas Schmitt wrote: > i have encrypted my HDD's (*) data partition. Now the disk access LED is > blinking rapidly as soon as i mount it. Could it possibly be the lazy init feature of ext4, which is enabled by default and can sometimes result in several minutes of background writes to a newly-created fs? https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=bfff68738f1cb5c93dab1114634cea02aae9e7ba https://www.thomas-krenn.com/en/wiki/Ext4_Filesystem#Lazy_Initialization Create with: mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 … to avoid this sort of thing. Cheers, Andy -- https://bitfolk.com/ -- No-nonsense VPS hosting
Riddling activity on encrypted and mounted partition
Hi, i have encrypted my HDD's (*) data partition. Now the disk access LED is blinking rapidly as soon as i mount it. Is this normal ? I did: cryptsetup -v -y luksFormat --type luks2 /dev/sda2 cryptsetup open /dev/sda2 daten dd if=/dev/zero bs=512 count=7679784591 status=progress of=/dev/mapper/daten Some hours later i did: mkfs -t ext4 /dev/mapper/daten In /etc/fstab i wrote a line with the new filesystem's UUID: UUID=... /daten ext4defaults,noauto Manual mounting mount /daten works fine, but also causes the disk activity. Unmounting ends it. The activity seems real: cat /sys/block/sda/sda2/stat ; sleep 10 ; cat /sys/block/sda/sda2/stat yields 797 117464313 3626 45984161 7636044108 7724077703 78712595 0 22193356 78716222000000 797 117464313 3626 45984321 7636044111 7724237503 78713043 0 22193688 78716669000000 which according to https://www.kernel.org/doc/Documentation/block/stat.txt shows activity with write fields: write I/Os 160 ( 45984321 - 45984161) write merges 3 (7636044111 - 7636044108) write sectors 159800 (7724237503 - 7724077703) write ticks448 ( 78713043 - 78712595) The disk was supposed to be idle in those 10 seconds. Its ext4 contains only 6 directories and 1 data file. Not much stuff to play with. The data partiton of my SSD, which holds the active system, shows much less activity during 10 seconds: write I/Os 7 write merges 1 write sectors 64 write ticks117 The disk is still the replaced WD WD4003FRYZ, now in a 5.25" noise reduction cage, which makes its periodic knocking sound unhearable unless i press my ear at the computer's side. The disk additionally holds a 64 GB swap partition. But swapoff /dev/sda1 does not influence the disk traffic on /dev/sda2. Only umount does. Have nice day :) Thomas