Bugzilla (bugzilla.redhat.com) will be under maintenance for infrastructure upgrades and will not be unavailable on July 31st between 12:30 AM - 05:30 AM UTC. We appreciate your understanding and patience. You can follow status.redhat.com for details.
Bug 1883464 - Hit error when shutdown 100 VMs and run `virsh domstats` concurrently
Summary: Hit error when shutdown 100 VMs and run `virsh domstats` concurrently
Keywords:
Status: NEW
Alias: None
Product: Red Hat Enterprise Linux Advanced Virtualization
Classification: Red Hat
Component: libvirt
Version: 8.3
Hardware: x86_64
OS: Linux
medium
medium
Target Milestone: rc
: 8.3
Assignee: Virtualization Maintenance
QA Contact: chhu
URL:
Whiteboard: libvirt_RHV_INT
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2020-09-29 09:53 UTC by chhu
Modified: 2020-11-05 19:46 UTC (History)
8 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed:
Type: Bug
Target Upstream Version:


Attachments (Terms of Use)
libvirtd.log (11.51 MB, application/gzip)
2020-09-29 09:59 UTC, chhu
no flags Details
qemu-guest.log (643.97 KB, application/gzip)
2020-09-29 10:00 UTC, chhu
no flags Details

Description chhu 2020-09-29 09:53:20 UTC
Description of problem:
Hit error when shutdown 100 VMs and run `virsh domstats` concurrently,
"error: internal error: End of file from qemu monitor
error: Unable to read from monitor: Connection reset by peer".

Version-Release number of selected component (if applicable):
libvirt-6.6.0-6.module+el8.3.0+8125+aefcf088.x86_64
qemu-kvm-5.1.0-9.module+el8.3.0+8182+ac9ced32.x86_64
kernel: 4.18.0-234.el8.x86_64

How reproducible:
100%

Steps to Reproduce:
1. Start 100 VMs
2. Run domstats.sh to get all domain stats on one terminal
# cat domstats.sh
#!/bin/sh
for i in {1..50}
do 
virsh domstats 2>&1 >>stats.log
sleep 2 
done

#sh domstats.sh

3. On another terminal, shutdown 100 VMs concurrently
# for i in $(virsh list --name); do virsh shutdown $i &  done

4. Get error in libvirtd.log and on the terminal which is running domstats.sh
-------------------------------------------------------------------
# sh domstats.sh
error: internal error: End of file from qemu monitor
error: Unable to read from monitor: Connection reset by peer
error: Unable to read from monitor: Connection reset by peer
error: Unable to read from monitor: Connection reset by peer
--------------------------------------------------------------------

Actual results:
In Step4, Get error in libvirtd.log

Expected results:
In Step4, No error.

Additional info:
- libvirtd.log
- qemu guest log

- In step2 run `python3 domstats.py`, hit same error.
# cat domstats.py 
import libvirt
import time
conn = libvirt.open("qemu:///system")
for x in range(1,100):
   print(conn.getAllDomainStats())
   time.sleep(2)

Comment 1 chhu 2020-09-29 09:59:50 UTC
Created attachment 1717488 [details]
libvirtd.log

Comment 2 chhu 2020-09-29 10:00:31 UTC
Created attachment 1717489 [details]
qemu-guest.log

Comment 4 Michal Privoznik 2020-10-12 07:06:40 UTC
Just for the record, I helped Chenli debug this and the problem is how libvirt deals with monitor disappearing during domstats. While we check whether monitor is available when getting stats for a domain, we do it upfront and don't take into account that it may disappear between two stat callbacks.


Note You need to log in before you can comment on or make changes to this bug.