Forwarded to devs CGroups disk write/read not working!

[B4c4rd1]

TITLE:

CGroups disk write/read not working!​

PRODUCT, VERSION, OPERATING SYSTEM, ARCHITECTURE:

Plesk Onyx, Version 17.5.3 Update #10, Ubuntu 16.04.2 LTS‬, x64​

PROBLEM DESCRIPTION:

After changing Disk Read and Disk Write have no effect.​

STEPS TO REPRODUCE:

Change the disk read or write to 10 MB/s

Show the following thread. It had worked: https://talk.plesk.com/threads/cgroups-disk-write-read-not-unlimited-after-a-change.341945/​

ACTUAL RESULT:

Without cache:

dd if=testfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 0,862584 s, 1,2 GB/s

With cache:

dd if=testfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 0,16022 s, 6,7 GB/s

EXPECTED RESULT:

1073741824 bytes (1,1 GB, 1,0 GiB) copied, 0,862584 s, 10 MB/s​

ANY ADDITIONAL INFORMATION:

root@srv-XXX ~ # plesk sbin resctrl --status
enabled
root@srv-XXX ~ # plesk sbin resctrl --features
{"cpu": true, "cpuacct": true, "memory": true, "blkio": true, "vhosts_blkio": true}

YOUR EXPECTATIONS FROM PLESK SERVICE TEAM:

Confirm bug​

 

[B4c4rd1]

I found my own mistake. The function can also only work if one limits the Ram and he actually has to read from the hard disk.

Edit: Now it does not work anymore. This should be checked.

Host Proxmox use ZFS
KVM use ext4

 

[B4c4rd1]

Hi Plesk Team,

Can you resolve this error?

Thanks!

 

[IgorG]

Developers can't reproduce and confirm this issue:

STR:

• Install Plesk 17.5.3 on Ubuntu 16.04 under KVM.
• Enable Resource controller
• Create a subscription
• Enable login to shell for subscription user
• Set the read limit to 10 Mb/s
• Login under root:
  
  
  root@a10-52-71-144:~# plesk version
  Product version: Plesk Onyx 17.5.3 Update #12
  Update date: 2017/06/29 09:57
  Build date: 2017/03/17 16:00
  OS version: Ubuntu 16.04
  Revision: 55d1b49a272f44666e1920eca8b6e4da449a38cd
  Architecture: 64-bit
  Wrapper version: 1.2
  root@a10-52-71-144:~# plesk sbin resctrl --status
  enabled
  root@a10-52-71-144:~# plesk sbin resctrl --features
  {"cpu": true, "cpuacct": true, "memory": true, "blkio": true, "vhosts_blkio": true}
  root@a10-52-71-144:~# mount | grep "on / type"
  /dev/vda5 on / type ext4 (rw,relatime,errors=remount-ro,data=ordered)
  
  
  
• Flush buffers under root:
  
  
  root@a10-52-71-144:~# echo 3 | sudo tee /proc/sys/vm/drop_caches
  3
  
  
  
• Login under subscription user.
• Create temprorary file:
  
  
  testsw@a10-52-71-144:~$ dd of=tempfile if=/dev/urandom bs=1M count=1024
  1024+0 records in
  1024+0 records out
  1073741824 bytes (1.1 GB, 1.0 GiB) copied, 51.1728 s, 21.0 MB/s
  
  
  
• Checking that the read limit is working:
  
  
  testsw@a10-52-71-144:~$ dd if=tempfile of=/dev/null bs=1M count=1024
  1024+0 records in
  1024+0 records out
  1073741824 bytes (1.1 GB, 1.0 GiB) copied, 102.375 s, 10.5 MB/s
  
  
  
• All is good.
• Change the read limit to 5 Mb/s
• Flush buffers under root:
  
  
  root@a10-52-71-144:~# echo 3 | sudo tee /proc/sys/vm/drop_caches
  3
  
  
  
• Checking that a read limit is working:
  
  
  testsw@a10-52-71-144:~$ dd if=tempfile of=/dev/null bs=1M count=1024
  1024+0 records in
  1024+0 records out
  1073741824 bytes (1.1 GB, 1.0 GiB) copied, 205.053 s, 5.2 MB/s
  
  
  
• All is ok.
• Change the read limit to Unlimited.
• Flush buffers under root:
  
  
  root@a10-52-71-144:~# echo 3 | sudo tee /proc/sys/vm/drop_caches
  3
  
  
  
• Checking that a read limit is working
  
  
  testsw@a10-52-71-144:~$ dd if=tempfile of=/dev/null bs=1M count=1024
  1024+0 records in
  1024+0 records out
  1073741824 bytes (1.1 GB, 1.0 GiB) copied, 204.809 s, 5.2 MB/s
  
  
  
• Test is fail -> Issue - CGroups disk write/read not unlimited after a change! still exists.

 

[B4c4rd1]

root@srv-XX ~ # plesk version
Product version: Plesk Onyx 17.5.3 Update #11
    Update date: 2017/06/26 18:10
     Build date: 2017/03/17 16:00
     OS version: Ubuntu 16.04
       Revision: 55d1b49a272f44666e1920eca8b6e4da449a38cd
   Architecture: 64-bit
Wrapper version: 1.2
root@srv-XX ~ # plesk sbin resctrl --status
enabled
root@srv-XX ~ # plesk sbin resctrl --features
{"cpu": true, "cpuacct": true, "memory": true, "blkio": true, "vhosts_blkio": true}
root@srv-XX ~ # mount | grep "on / type"
/dev/mapper/srv--XX--vg-root on / type ext4 (rw,relatime,errors=remount-ro,data=ordered,jqfmt=vfsv1,usrjquota=aquota.user,grpjquota=aquota.group)
root@srv-XX ~ # echo 3 | sudo tee /proc/sys/vm/drop_caches
3

On subscription:

Write Limit: 10 MB/s

XXX@srv-XX / $ dd of=tempfile if=/dev/urandom bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 66,3239 s, 16,2 MB/s

Read Limit: 10 MB/s (Delete Cache before)

XXX@srv-XX / $ dd if=tempfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 102,403 s, 10,5 MB/s

Write Limit: 5 MB/s

XXX@srv-XX / $ dd of=tempfile if=/dev/urandom bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 66,1781 s, 16,2 MB/s

Read Limit: 5 MB/s (Delete Cache before)

XXX@srv-XX / $ dd if=tempfile of=/dev/null bs=1M count=1024
1024+0 records in
1024+0 records out
1073741824 bytes (1,1 GB, 1,0 GiB) copied, 204,9 s, 5,2 MB/s

What is with Disk Write limit?

 

[IgorG]

I would suggest you submit ticket to Plesk Support Team. Experienced supporters will check this issue directly on your server.

 

[B4c4rd1]

The problem I have on several servers. Also on a clean test system. The writing speed can not be throttled.