Sqlyog Job Agent Keeps Running

[uyuni]

Sometimes SJA doesn't exit normally but keeps on running. I have a task in the windows scheduler and it stays “running”. Because of this it isn't started again.

Some dump files SJA_Dump_00x.dmp were created but they are emtpy.

What can I do to give you more information so you can reproduce this?

Emilio

[peterlaursen]

An empty DUMP file is created when the program starts. The file is removed when it closes. If the program is 'killed' (not shutting down gracefully) the file will remain. Also it will remain as long as the job is running.

Please tell (for a start)

* Windows version

* remote or local servers? How do you connect (HTTP, SSH, direct)

* what kind of SJA job? data sync?

* how long does it normally take to complete the job?

* how can you be sure that job has completed?

[uyuni]

peterlaursen wrote on Oct 9 2007, 10:33 AM:

An empty DUMP file is created when the program starts. The file is removed when it closes. If the program is 'killed' (not shutting down gracefully) the file will remain. Also it will remain as long as the job is running.

Please tell (for a start)

* Windows version

* remote or local servers? How do you connect (HTTP, SSH, direct)

* what kind of SJA job? data sync?

* how long does it normally take to complete the job?

* how can you be sure that job has completed?

Hello,

– OS: Server 2003

– connection to remote SSH server

– task: a maintenance task, an UPDATE on the remote server

– normal completion in 6 seconds

– I cannot check if the task has completed correctly because the database has changed since then.

Emilio

[peterlaursen]

“- connection to remote SSH server”

I think the problem may be that connection between PLINK and the SSH daemon gets broken. And SJA does not get 'lost connection' from the compiled-in client API.

So SJA simply may just be waiting …

A similar issue has been confirmed for Vista, but maybe that also appiles to 2003!

MySQL bug report: http://bugs.mysql.com/bug.php?id=31109

We should test that, but you are welcome too!

[uyuni]

For such errors it would be nice to have a timeout for the SJA job (or maybe only for an SQL statement). Configuration of the timeout as a parameter for the commandline would be sufficient.

Now my whole system breaks because never a new sja is started from the scheduler until I kill the old sja.

Emilio

[peterlaursen]

I understand the problem. I can only say that we need to analyze what we can do.

Do you have any indication that its is such thing that is happening for you?

[uyuni]

I don't have a real indication but this seems very probable to me.

[uyuni]

The same happened again now and a process PLINK was also running!

[peterlaursen]

a PLINK process will continue to run if SQLyog or SJA is running using SSH tunnelling, and the 'mother program' (SQLyog, SJA) is not closing down 'gracefully' (ie: killed or crashing). This PLINK process is now 'orphaned' and it will run untill it is killed or system reboots. I do not think it can be otherwise on Windows.

When closing normally SQLyog and SJA will close the PLINK process too.

I will add the timeout request to our issue-tracker today!

[peterlaursen]

We will have to FULLY understand this before deciding what to do!

For some reason you have experienced that sometimes a SJA job runs for very long time (seems like 'running forever'). However this will NOT prevent another SJA job to start. You can have hundreds of SJA jobs running at the same time (provided that the server allows for that number of simultaneous connections).

I think the problem is SSH-tunnelling! A SSH connection starts a PLINK instance. That PLINK instance ocupies a local port. That port will be unavailable as long as this PLINK process has not been closed by its 'parent' SJA process (or explicitly killed by a human user). An attempt to start another job with the same connection settings (including the same local port for SSH) will fail, because the port is not available to the new job.

Can you confirm this is what happens?

Actually I think we should consider to make it possibe specify a port RANGE. Let's says port range 3310-3350 was specified. If connnection fails on 3310 it will next attempt 3311, and next 3312 and so on untill the end of the specified range.

That does not exclude that also a setting for a SJA job could also be a good idea!

[Author]

Posts