nagios-plugins 1.4.15 check_procs broken, fixed on Sourceforge

[sonikbuddha@…]

The 1.4.15 version of nagios-plugins contains a broken check_procs due to a wrong ps command, listed below. I've built from the version on Sourceforge (frustratingly with the same release number) and the bug is fixed. Please upgrade to the version on Sourceforge.

​http://sourceforge.net/projects/nagiosplug/files/nagiosplug/

Problem: The macports released version of 1.4.15 containing check_procs runs this ps command:

CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu etime comm args'

The 'comm' argument does not present the command without the path, causing all command searches (check_procs -C <command>) to fail. In this release this plug-in will never process command searches properly.

The fixed version, on Sourceforge now, runs this command:

CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'

The 'ucomm' properly presents the command without the path and allows the plug-in to properly process a '-C'. I'm presenting this as a priority fix as a) its simply a package upgrade, not even a version change b) as is, this monitoring plug-in presents incorrect information.

For those seeing a temporary work around, the nagios-plugins source package on Sourceforge builds natively on at least Snow Leopard. OR run check_procs -a <command>, which has similar functionality.

[sonikbuddha@…]

Cc Me!

[jmroot (Joshua Root)]

Please remember to cc the maintainer. You do not need to be in cc when you are the reporter.

[markd@…]

We don't use snapshots in macports normally. Can you do a diff of the check_procs.c between the released tar file and the git version and submit it? Instructions for making patches are in the guide.

[markd@…]

I found the patch from the git repository for check_procs.c and applied it in r76989. Please test and report if it fixes it.

[sonikbuddha@…]

I've upgraded to 1.4.15_2 and I'm still getting the broken ps command.

check_procs -C bash -vvv CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu etime comm args'

[markd@…]

Well I can't fix it if you can't tell me what fixed it for you.

[sonikbuddha@…]

When I compiled from the source available on Sourceforge's download site it worked for me; I have not tried the source in the nagios-plugin git repository. ​http://sourceforge.net/projects/nagiosplug/files/nagiosplug/1.4.15/

Could your experience suggest the problem is upstream?

[markd@…]

Oh I'm sorry, I understand now. I just realized the problem must be in the configure switches. Do a 'port ed nagios-plugins' (if your EDITOR variable is set) and you'll see these options. Tell me what mods need to be made.

--with-ps-command="/bin/ps -axwo 'stat uid pid ppid vsz --with-ps-format="%s %d %d %d %d %d %f %s %s %n" \ --with-ps-cols=10 \ --with-ps-varlist="procstat,&procuid,&procpid,&procppid,

[sonikbuddha@…]

In my configure, I passed no arguments. Its interesting that the source package that you've pulled does not have the same defaults. I spelunked a bit through what I had and found this. Please be patient, I rarely dig into the build process or have troubleshoot issues beyond passing configure arguments.

From my local build that I made from the Sourceforge package, my config.log lists | #define PS_VARLIST procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos | #define PS_COMMAND "/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'" | #define PS_FORMAT "%s %d %d %d %d %d %f %s %n" | #define PS_COLS 9

And the config.h: /* Number of columns in ps command */ #define PS_COLS 9

/* Verbatim command to execute for ps in check_procs */ #define PS_COMMAND "/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'"

/* Format string for scanning ps output in check_procs */ #define PS_FORMAT "%s %d %d %d %d %d %f %s %n"

/* Whether the ps utility uses the "procetime" field */ /* #undef PS_USES_PROCETIME */

/* Variable list for sscanf of 'ps' output */ #define PS_VARLIST procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos

Taking your advice, based on this, I changed

--with-ps-command="/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu etime comm args'" \ --with-ps-format="%s %d %d %d %d %d %f %s %s %n" \ --with-ps-cols=10 \ --with-ps-varlist="procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procetime,procprog,&pos"

to

--with-ps-command="/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'" \ --with-ps-format="%s %d %d %d %d %d %f %s %n" \ --with-ps-cols=9 \ --with-ps-varlist="procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos"

And rebuilt.

The result worked: check_procs -C bash -vvv CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command' ... PROCS OK: 1 process with command name 'bash'

Just because I was worried that there was no configure option for proctime (which I removed from the ps args list), I compared my config.h to the macports config.h I just generated and the lines match. I can share the config.log/config.h/Makefile, if necessary.

I hope this helps.

[sonikbuddha@…]

Ah, missed the wiki formatting. More readable:

From my local build that I made from the Sourceforge package, my config.log lists

| #define PS_VARLIST procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos
| #define PS_COMMAND "/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'"
| #define PS_FORMAT "%s %d %d %d %d %d %f %s %n"
| #define PS_COLS 9

And the config.h:

/* Number of columns in ps command */
#define PS_COLS 9

/* Verbatim command to execute for ps in check_procs */
#define PS_COMMAND "/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'"

/* Format string for scanning ps output in check_procs */
#define PS_FORMAT "%s %d %d %d %d %d %f %s %n"

/* Whether the ps utility uses the "procetime" field */
/* #undef PS_USES_PROCETIME */

/* Variable list for sscanf of 'ps' output */
#define PS_VARLIST procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos

Taking your advice, based on this, I changed

--with-ps-command="/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu etime comm args'" \
--with-ps-format="%s %d %d %d %d %d %f %s %s %n" \
--with-ps-cols=10 \
--with-ps-varlist="procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procetime,procprog,&pos" 

to

--with-ps-command="/bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'" \
--with-ps-format="%s %d %d %d %d %d %f %s %n" \
--with-ps-cols=9 \
--with-ps-varlist="procstat,&procuid,&procpid,&procppid,&procvsz,&procrss,&procpcpu,procprog,&pos"

And rebuilt.

The result worked:

check_procs -C bash -vvv
CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'
...
PROCS OK: 1 process with command name 'bash'

The rest you can gather from above (next time I'll preview)

[markd@…]

Okay, I made the changes. Thank you very much @sonikbuddha.

You know, I originally did not specify any --with-ps config switches, and someone made a ticket and told me to make them as they were. This raises the question: should the --with-ps- switches be specified at all, or should they be omitted and just let the defaults be used? What do you think would be best?

[markd@…]

Changed in r77000.

[sonikbuddha@…]

Cleaned and upgraded to 1.4.15_3, tested, and worked!

check_procs -C bash -vv
CMD: /bin/ps -axwo 'stat uid pid ppid vsz rss pcpu ucomm command'
Matched: uid=1043 vsz=2435548 rss=1076 pid=47905 ppid=47899 pcpu=0.00 stat=Ss etime= prog=bash args=-bash
Matched: uid=1043 vsz=2435548 rss=1140 pid=35774 ppid=35773 pcpu=0.00 stat=S etime= prog=bash args=bash
Matched: uid=1043 vsz=2435548 rss=1128 pid=36368 ppid=36367 pcpu=0.00 stat=S+ etime= prog=bash args=bash
Matched: uid=1043 vsz=2435548 rss=1140 pid=36417 ppid=36416 pcpu=0.00 stat=S+ etime= prog=bash args=bash
PROCS OK: 4 processes with command name 'bash'

If the defaults are the same in the source package you have and the package I downloaded from Sourceforge I'd go with those since they, well, work. That is, based upon this experience. I can't speak to the reasoning behind the prior ticket.

Thanks for your quick responses!