cbsd jexec jname="*" date .... hangs.

[tm-up]

Running above command will hang (all related pkg commands too).
I see two related processes, while the command hangs:

/usr/local/bin/sudo -E /usr/local/cbsd/sudoexec/jexec jname=* date
/usr/local/bin/cbsd /usr/local/cbsd/sudoexec/jexec jname=* date

Both processes are marked stopped ('T' in ps -aux).
Output on console:

# cbsd jexec jname='*`
Hint: Press 'Ctrl+t' to see last logfile line for active task
 * [jexec: pending: 2:[ myjail testjail], active: 0]

Running the command cbsd jexec jname=myjail date works.

FreeBSD version: 12.0 RELEASE p11
CBSD version: 12.1.1

[tm-up]

running that command with env CBSD_DEBUG=1 cbsd jexec jname='*' date produces a lot of outputs, the last line being:

+ task 'mode=new' 'logfile=/usr/jails/tmp/jexec_multiple.myjail.log.61670' 'client_id=myjail' 'autoflush=0' 'owner=jexec_multiple' /usr/bin/env 'NOCOLOR=1' /usr/local/bin/cbsd jexec 'jname=myjail' date

The mentioned log file does not exist.

[tm-up]

I just observe that the cbsd process consumes approx. 25% CPU time (info from "top"- command).

The command cbsd jstart jname='*' or the equivalent jstop command works as expected.
UPDATE: I am not so sure whether this is completely true: The last three jails get started, but three processes are left over and their status 'Is': daemon: /usr/local/bin/cbsd[14806] (daemon).
The process 14806 is "<defunct>". (same for the other two daemon processes of cbsd).
When stopping those jails, I'll get the start-console output and then the jails stop.
There are more zombie processes belonging either to a "nice" command issued by CBSD or the above mentioned "daemon".

Any ideas what is going on here?

[olevole]

can you post output for:

cbsd taskls

?

I also have an idea to rewrite 'cbsd task mode=new' and cbsdd with beanstalkd-based engine (minimalist broker, was demonstrated in https://www.bsdstore.ru/en/broker_driven_sample_ssi.html ).

However, the current implementation should handle simple operations like this.

[olevole]

Currently, all parallel/massive commands work according to the following principle:

1. add cmd to queue,e.g:

cbsd task mode=new date

2. cbsdd process (daemonized shell/cbsd, /usr/local/etc/rc.d/cbsdd) watch via kqueue(2) for taskd.sqlite modification ( ~cbsd/var/db/cbsdtaskd.sqlite ) and wakeup when any new task is coming.

3. cbsdd execute commands with output to /tmp/*.log and update result/status.

4. jexec/jstart/jstop/... command just 'poll' job status and read /tmp/*.log when status 'complete'

[tm-up]

The problem here was simple and stupid: cbsdd was not running...