Bug 2229441

Summary: [abrt] pcp: fsync(): pmlogconf killed by SIGABRT
Product: [Fedora] Fedora Reporter: bdp <inbox1>
Component: pcpAssignee: Nathan Scott <nathans>
Status: ASSIGNED --- QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 38CC: agerstmayr, inbox1, jkurik, lchilton, nathans, sfeifer
Target Milestone: ---Keywords: Bugfix, Triaged
Target Release: ---   
Hardware: x86_64   
OS: Unspecified   
URL: https://retrace.fedoraproject.org/faf/reports/bthash/bddb3ce75bb9ee716d773743f5a7a70efe9c845
Whiteboard: abrt_hash:0f632ab6decc762f820cfdabf59fd0de93852d22;VARIANT_ID=workstation;
Fixed In Version: Doc Type: No Doc Update
Doc Text:
If this bug requires documentation, please select an appropriate Doc Type value.
 Story Points: ---
Clone Of: Environment:
Last Closed: Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
File: proc_pid_status
none
File: maps
none
File: limits
none
File: environ
none
File: open_fds
none
File: mountinfo
none
File: os_info
none
File: cpuinfo
none
File: core_backtrace
none
File: dso_list
none
File: backtrace none

Description bdp 2023-08-06 01:50:44 UTC 
Version-Release number of selected component:
pcp-6.0.5-1.fc38

Additional info:
reporter:       libreport-2.17.11
type:           CCpp
reason:         pmlogconf killed by SIGABRT
journald_cursor: s=e882cb7b29f74a45837e2237194d5204;i=132eeed;b=bbc88af5bd53490faf48286d0fe1f378;m=63ade593;t=601d9b0fc347f;x=ddc19a54a15f501b
executable:     /usr/bin/pmlogconf
cmdline:        /usr/libexec/pcp/bin/pmlogconf -r -c -q -h local: /var/lib/pcp/config/pmlogger/config.default.tmp
cgroup:         0::/system.slice/pmlogger.service
rootdir:        /
uid:            0
kernel:         6.4.7-200.fc38.x86_64
package:        pcp-6.0.5-1.fc38
runlevel:       N 5
backtrace_rating: 4
crash_function: fsync

Truncated backtrace:
Thread no. 1 (2 frames)
 #0 fsync at ../sysdeps/unix/sysv/linux/fsync.c:27
 #1 pmlogconf at /usr/src/debug/pcp-6.0.5-1.fc38.x86_64/src/pmlogconf/pmlogconf.c:1772
Comment 1 bdp 2023-08-06 01:50:47 UTC 
Created attachment 1981898 [details]
File: proc_pid_status
Comment 2 bdp 2023-08-06 01:50:49 UTC 
Created attachment 1981899 [details]
File: maps
Comment 3 bdp 2023-08-06 01:50:50 UTC 
Created attachment 1981900 [details]
File: limits
Comment 4 bdp 2023-08-06 01:50:51 UTC 
Created attachment 1981901 [details]
File: environ
Comment 5 bdp 2023-08-06 01:50:52 UTC 
Created attachment 1981902 [details]
File: open_fds
Comment 6 bdp 2023-08-06 01:50:53 UTC 
Created attachment 1981903 [details]
File: mountinfo
Comment 7 bdp 2023-08-06 01:50:55 UTC 
Created attachment 1981904 [details]
File: os_info
Comment 8 bdp 2023-08-06 01:50:56 UTC 
Created attachment 1981905 [details]
File: cpuinfo
Comment 9 bdp 2023-08-06 01:50:57 UTC 
Created attachment 1981906 [details]
File: core_backtrace
Comment 10 bdp 2023-08-06 01:50:58 UTC 
Created attachment 1981907 [details]
File: dso_list
Comment 11 bdp 2023-08-06 01:50:59 UTC 
Created attachment 1981908 [details]
File: backtrace
Comment 12 Nathan Scott 2023-08-07 06:10:01 UTC 
Hi there,

Thanks for reporting the issue!  I've been digging into it and have not found any coding issue in pmlogconf itself, other than a small, unrelated memory leak.  I'm suspecting the fsync is inducing a service (re)start delay due to background I/O activity, and that this causes a systemd timeout to be exceeded and SIGABRT sent.

We don't have a particularly strong need to be fsync'ing in pmlogconf, so as an initial step I'll drop that call - at the very least this will make bootup slightly faster for the pmlogger and pmlogger_farm services.

This is being tracked via upstream github PR now and will be included in the next PCP release, which will reach Fedora in a few weeks time:
https://github.com/performancecopilot/pcp/pull/1789