CPU load, because service is hanging

16 views
Skip to first unread message

Eirik Zakariassen

unread,
Jan 15, 2026, 4:38:15 AM (10 days ago) Jan 15
to Event-Driven Servers
Hi again Marc.

I've noticed something interesting lately where CPU load spikes on the nodes running tacplus, it seems like when the service spawns additional workers, they just hang in a ~50-100% load state and never die properly, main service still runs, but machine gets sluggish and I get warnings in vcenter.

If I kill those hanging services it is fine.

Seen anything like that before ?

top - 10:31:13 up 19:08,  2 users,  load average: 5.87, 5.47, 5.23
Tasks: 333 total,   7 running, 326 sleeping,   0 stopped,   0 zombie
%Cpu(s): 17.3 us, 82.1 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.3 hi,  0.2 si,  0.0 st
MiB Mem :  15732.3 total,   8310.1 free,   1976.6 used,   6090.2 buff/cache
MiB Swap:   8192.0 total,   8192.0 free,      0.0 used.  13755.7 avail Mem

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND  
    977 root      20   0   21276  11404   7432 R  83.1   0.1 901:05.61 tac_plu+
   1954 root      20   0   23640  11472   7740 R  77.4   0.1 900:55.76 tac_plu+
   1990 root      20   0   21416  11180   7744 R  77.1   0.1 901:16.74 tac_plu+
   2062 root      20   0   21276  11712   7768 R  76.7   0.1 899:44.29 tac_plu+
   1981 root      20   0   23640  11436   7736 R  75.4   0.1 901:18.51 tac_plu+
   1072 root      20   0  680520  56924  54672 S   1.7   0.4  10:55.39 rsyslogd
  60427 root      20   0   21272  11340   7912 S   1.0   0.1   0:00.56 tac_plu+
    730 root      20   0  137672  65164  63668 S   0.3   0.4   2:18.17 systemd+
  60177 root      20   0       0      0      0 I   0.3   0.0   0:00.41 kworker+
      1 root      20   0  174744  17416  11252 S   0.0   0.1   0:54.41 systemd  
      2 root      20   0       0      0      0 S   0.0   0.0   0:00.02 kthreadd 


● tac_plus.service - TACACS+ Service
     Loaded: loaded (/etc/systemd/system/tac_plus.service; enabled; preset: dis>
     Active: active (running) since Wed 2026-01-14 15:26:48 CET; 19h ago
   Main PID: 2191 (tac_plus-ng)
      Tasks: 6 (limit: 100406)
     Memory: 839.9M (peak: 2.2G)
        CPU: 3d 3h 12min 46.061s
     CGroup: /system.slice/tac_plus.service
             ├─ 2191 "tac_plus-ng: 0 connections, accepting up to 300 more"
             ├─60427 "tac_plus-ng: 0 connections"
             ├─60428 perl /usr/local/lib/mavis/mavis_tacplus_ldap.pl
             ├─60429 perl /usr/local/lib/mavis/mavis_tacplus_ldap.pl
             ├─60430 perl /usr/local/lib/mavis/mavis_tacplus_ldap.pl
             └─60431 perl /usr/local/lib/mavis/mavis_tacplus_ldap.pl

Jan 15 10:30:12 hostname.domain tac_plus-ng[59831]: 10.0.0.0 >
Jan 15 10:30:12 hostname.domain tac_plus-ng[59831]: 10.0.0.0 >
Jan 15 10:30:12 hostname.domain tac_plus-ng[59831]: 10.0.0.0 >
Jan 15 10:30:12 hostname.domain tac_plus-ng[59831]: 10.0.0.0 >
Jan 15 10:30:16 hostname.domain tac_plus-ng[59831]: - Exiting.
Jan 15 10:30:16 hostname.domain tac_plus-ng[60423]: - Exiting.
Jan 15 10:30:16 hostname.domain tac_plus-ng[2191]: scm_send_msg: s>
Jan 15 10:30:16 hostname.domain tac_plus-ng[2191]: scm_send_msg (/>

Marc Huber

unread,
Jan 17, 2026, 3:59:29 AM (8 days ago) Jan 17
to event-driv...@googlegroups.com

Hi Eirik,

that obviously shouldn't happen. Could you please share your "tac_plus-ng -v" output?

Thanks,

Marc

--
You received this message because you are subscribed to the Google Groups "Event-Driven Servers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to event-driven-ser...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/event-driven-servers/2d908953-be09-4579-9f0a-bb40e3844b74n%40googlegroups.com.

Eirik Zakariassen

unread,
Jan 17, 2026, 2:08:20 PM (8 days ago) Jan 17
to Event-Driven Servers
Hi Marc.

This is the output

/usr/local/sbin/tac_plus-ng -v
tac_plus-ng version 7bc67dc7f721abd73106826d89c96f49acd63aea/PCRE2/ARES 1.19.1/CURL 7.76.1/OpenSSL 3.5.1 1 Jul 2025

Marc Huber

unread,
Jan 17, 2026, 2:19:13 PM (8 days ago) Jan 17
to event-driv...@googlegroups.com

Hi Eirik,

that GIT hash time stamp is Sun Nov 2 10:07:40 2025 +0100 -- please update to the current GIT, I'm pretty sure that that that the issue you're seeing is already fixed.

In case that busy loop still persists: you can uncomment the "#define DEBUG_CB_STATS" in misc/io_sched.h and rebuild, that enables debug stats which might give some insight on the callbacks involved.

Cheers,

Marc

Eirik Zakariassen

unread,
Jan 19, 2026, 3:18:06 AM (6 days ago) Jan 19
to Event-Driven Servers
Thanks Marc.

Did an upgrade today

tac_plus-ng version ab4b5d2594e21984aa35a30bfc7a069cd430c2b4/PCRE2/ARES 1.19.1/CURL 7.76.1/OpenSSL 3.5.1 1 Jul 2025

I'll keep an eye on it.
Reply all
Reply to author
Forward
0 new messages