QMC Task Error: "No slave scheduler found"

[daniel_dalnekoff]

Hi,

We are getting an error in the QMC when scheduled tasks are starting to queue up (waiting on what is executing to begin).  My timeout is set to 120 minutes, but these are immediately failing with the error above (No slave scheduler found).

We have a single scheduler node configured as Master and Slave, and the engine service is running on this node.  Confirmed QMC shows the service is running (and so does Windows).  Resources on the machine are not maxed out.  I am setup to allow 6 concurrent tasks to run (based on 8 cores).  It seems like as soon as several tasks start to get in the triggered or queued status this happens.

Does anyone have any thoughts on how to troubleshoot?

[plexpro52]

Hello Dan et al.,

Our environment is similar to yours, a single node, master and slave, etc.  I have a daily task which loads a number of qvd files, and then five app loads chained to its success.  A week or two after installing the June release, it started to behave abnormally, three of the five (not predictable as to which three) would fail with the error that you have noted.

Qlik provided us with an idea to try, which worked for us.

They had us add a line to the C:\ProgramData\Qlik\Sense\Engine\settings.ini file.  In the [Settings 7] section, we added "DisableNewRowApplicator=0", and that resolved the problem, at least on first execution of the job stream.  We'll see if it sticks.

René

[BeeGees]

Thx for share. Seems worth to try it.

My workaround was harsh and simple. Since tasks failed randomly the solution was just to automate their retry. Therefore i wrote an SQL query which allows me to define those tasks that failed:

        SELECT

            "ReloadTasks"."ID",

            "ReloadTasks"."Name"

        FROM

            public."ExecutionResults"

            join public."ReloadTaskOperationals" on "ReloadTaskOperationals"."LastExecutionResult_ID" = "ExecutionResults"."ID"

            join public."ReloadTasks" on "ReloadTasks"."ID" = "ExecutionResults"."TaskID"

        where

            "Status" = 11

        order by "ExecutionResults"."StartTime" asc

Then I just rerun the task using 3rd party tool QlikSenseTask.exe Works as a charm.

Above is scheduled to repeat every 1 min. I realise that could be possible done using API, but it was faster for me that way.

[Daniele_Purrone]

Hi René,

thanks for informing all.
May I ask you who you talked with at Qlik? Was this a support case? We are trying to investigate if this is caused by a bug.

Thanks!

Daniele

[plexpro52]

Hello Daniele,

>> May I ask you who you talked with at Qlik?

Analytics8 is our level 1 support.  They worked with the Qlik support organization, who provided this solution.

>> Was this a support case?

Yes, it was a support case with Analytics8.  I don't know how Qlik handled it at their end.

René

[Daniele_Purrone]

Thanks René, I've found the support case.

We are currently investigating why this happens.
In your situation it was easy to find a workaround, because of the presence of AAALR warnings in the logs.
We tried to apply to same workaround to other customers, and it worked in some cases (even without AAALR warnings being present), it didn't in some other.
The issue might be connected to high cpu usage.

As said, Support is currently investigating this. Anyone having the problem, please try the workaround first, and should not that help, feel free to reach out to Qlik Support (either directly or via the responsible partner, if you have one).

Thanks in advance!

[peter_turner]

Hello,

I support a number of customers and have encounted this exact behaviour on at least three seperate installations.

I'll try and list the findings i have so far, and if i manage to resolve the problems...

Servers are running either April or June 2018 (inital release and Patch1) QS, some single and some multinode.

The problem is the best kind, intermittent.

It appears on the QMC task as 'Failed' status, with the information 'No slave scheduler found'.

Tasks will then run fine when manually triggered.

To me it seems to happen more when there are multiple tasks dependent on another task, maybe an issue with the tasks all being triggered together and not being able to wait afew seconds for an engine to become available?

In all cases the servers have lots of available RAM/CPU, scheduler service is running fine as reported by the OS and QMC.

The servers are not overloaded with Tasks.

I have tried adjusting the max concurrent reloads up and down, no change.

On the multinode servers i have adjusted the scheduler service role (between Master and Slave, Master only and Slave only) no change.

Restarting the Qlik Scheduler Service will resolve the issue for a while, depending on the frequency of reloads.

[BeeGees]

I can confirm 100% with Your observations. I did not post a ticket on support, since my workaround works perfectly (and i do not wanna waste my team with calls with support anymore, anyway), but my case is exactly as You described it. Therefore could You please post here in which version of Qlik it will be fixed (casue i assume it will be marked as a bug and fixed in future releases/patches)?

[Daniele_Purrone]

Thanks Peter!

[Daniele_Purrone]

I will. Just remind me if you don't hear from me or someone else at support in the next couple of weeks, please!

[Daniele_Purrone]

Peter, if you haven't done it already, please log a support case for every customer environment. This will help us gain some traction. Thanks!