| BUG: "prlctl destroy <UUID>" autostarts containers [message #53710] |
Wed, 18 November 2020 21:13  |
mstauber
Messages: 7
Registered: April 2009
|
Junior Member |
|
|
Hello,
I just ran into a nasty "split-brain" situation because of something that appears a bug. \o/
Situation:
Two nodes, "master" and "slave". Both have the same CTs. "master" actively runs them, while "slave" has a copy of each CT for backup purposes.
Relevant versions involved:
[root@slave ~]# cat /etc/redhat-release
Virtuozzo Linux release 7.8
[root@slave ~]# prlctl --version
prlctl version 7.0.201
[root@slave ~]# uname -a
Linux master.xxx.xxx 3.10.0-957.12.2.vz7.96.21 #1 SMP Thu Jun 27 15:10:55 MSK 2019 x86_64 x86_64 x86_64 GNU/Linux
Yes, both nodes haven't been rebooted in a few months, so the running kernel is a bit older. Mea culpa. Uptime is king.
On "slave" I needed to remove an older CT. Which was stopped. In fact all CTs on "slave" were stopped AND had "autostart" off:
[root@slave ~]# prlctl list --info|grep Autostart
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
So I ran "prlctl destroy ff717715-916c-497d-b9b5-bfab7cf8e3b5" and the CT in question was gone.
All good? Hell no! Take a look at this:
[root@slave ~]# cat /var/log/vzctl.log|grep "2020-11-18"|egrep "Start|destroyed"
2020-11-18T12:13:54-0600 vzctl : CT ff717715-916c-497d-b9b5-bfab7cf8e3b5 : Container private area was destroyed
2020-11-18T12:14:40-0600 vzctl : CT e8f386c2-08b5-45b9-b3ec-74e825ab082a : Starting Container ...
2020-11-18T12:14:40-0600 vzctl : CT f539028b-89a3-4f48-ad3a-6ea123e84d68 : Starting Container ...
2020-11-18T12:14:40-0600 vzctl : CT 48073f02-aeb7-4be8-911a-94daf2ce410d : Starting Container ...
2020-11-18T12:14:40-0600 vzctl : CT 3e2b0ace-604f-4297-9136-66f0cd0b48dc : Starting Container ...
2020-11-18T12:14:40-0600 vzctl : CT 4d615464-229e-4bfd-9b99-638563d116de : Starting Container ...
[...]
So *right* after destroying ff717715 as desired, without being asked to do so OpenVZ started up all *other* containers. Despite them all having 'autostart' off.
And of course the CTs on "slave" won the ARP-roulette and started serving traffic, which caused a hell of a mess.
I'm sorry, but why does deletion of a CT cause OpenVZ to fire up all stopped CTs even if they have "autostart" off? That's HIGHLY unexpected behavior.
With best regards,
Michael Stauber
|
|
|
|
| Re: BUG: "prlctl destroy <UUID>" autostarts containers [message #53711 is a reply to message #53710] |
Thu, 19 November 2020 07:09  |
khorenko
Messages: 533
Registered: January 2006
Location: Moscow, Russia
|
Senior Member |
|
|
Hi Michael,
i failed to reproduce the issue in an easy way:
[root@sbf7 ~]# prlctl list --info|grep Autostart
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
Autostart: off
[root@sbf7 ~]# prlctl list -a
UUID STATUS IP_ADDR T NAME
{e184beed-db92-4b07-971d-e8a4c9bff9f2} stopped - CT 100
{49bd6422-d6e7-4777-baaa-baa908e93f91} stopped - CT 101
{99b797a2-9b50-4a60-a9d7-30d3268c8ed9} stopped - CT 102
{863f365c-bcf0-45b0-84e8-229fac815129} stopped - CT 1024
{e0860e37-19f1-4c97-9374-343244f476fa} stopped - CT 200
{24f9c50f-f9a4-4784-b9da-4975363665fd} stopped 192.169.9.89 CT 300
{a6fb4d5a-2adb-443c-a01b-c12b12becb29} stopped 192.169.9.90 CT 301
{bd7900ba-a98d-47ae-b232-4ffd0e12e3b0} stopped - CT 302
{fbb33119-82a4-40a7-9d4e-dd28c3a3ee3e} stopped - CT 500
[root@sbf7 ~]# prlctl destroy fbb33119-82a4-40a7-9d4e-dd28c3a3ee3e
Removing the CT...
The CT has been successfully removed.
[root@sbf7 ~]# prlctl list -a
UUID STATUS IP_ADDR T NAME
{e184beed-db92-4b07-971d-e8a4c9bff9f2} stopped - CT 100
{49bd6422-d6e7-4777-baaa-baa908e93f91} stopped - CT 101
{99b797a2-9b50-4a60-a9d7-30d3268c8ed9} stopped - CT 102
{863f365c-bcf0-45b0-84e8-229fac815129} stopped - CT 1024
{e0860e37-19f1-4c97-9374-343244f476fa} stopped - CT 200
{24f9c50f-f9a4-4784-b9da-4975363665fd} stopped 192.169.9.89 CT 300
{a6fb4d5a-2adb-443c-a01b-c12b12becb29} stopped 192.169.9.90 CT 301
{bd7900ba-a98d-47ae-b232-4ffd0e12e3b0} stopped - CT 302
So please file a bug at bugs.openvz.org and provide full logs: vzctl.log and prl* logs.
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
