Dll Library was found but could not be loaded in FME-CodePudding

I'm having some issues with our FME Server which I don't understand. Every now and then, when running a nightly scheduled job, a FeatureReader fails to read a CSV file. The log returns:

1.  Trying to find a DYNAMIC plugin for reader named `CSV2'
2.  Library 'D:\Program Files\FMEServer\Server\fme\plugins/CSV2.dll' was found but could not be loaded. Ensure that all the dependent modules exist for this library
3.  Module 'CSV2' is unavailable for use with this FME edition
When I resubmit the job, everything works fine... which is great for the data being processed, but harder to debug.

The issue feels somewhat familiar and I did find similar questions but not related to CSV2.dll. I'm quite sure this is not related to desktop/server versions. Version is 2019.1.2. As far as I am aware nothing has changed in the installation, I do know the certificates were replaced recently but I doubt this has anything to do with it.

Not sure where I should start looking, can anyone point me in the right direction?

I found some jobs with the same kind of logging. These are writers instead of readers and caused the job to fail.

1.  Trying to find a DYNAMIC plugin for writer named `ESRIJSON'
2.  Library 'D:\Program Files\FMEServer\Server\fme\plugins/JSON.dll' was found but could not be loaded. Ensure that all the dependent modules exist for this library
3.  Module 'ESRIJSON' is unavailable for use with this FME edition

and

1.  Trying to find a DYNAMIC plugin for writer named `CSV'
2.  Library 'D:\Program Files\FMEServer\Server\fme\plugins/CSV.dll' was found but could not be loaded. Ensure that all the dependent modules exist for this library
3.  Module 'CSV' is unavailable for use with this FME edition

CodePudding user response：

The cause of this error is memory related.

In our case this happened during the night when several update jobs are scheduled. We have a bunch of schedules with the same start time and let the queue work through the jobs.

A possible solution was to lower the number of succesful jobs to initiate an engine restart.

To check if this might be the issue, I re-scheduled the jobs with issues an hour later and the problem went away.

This issue also might be caused by a server / desktop version mismatch, desktop being newer than server. In our case this was not a possible cause.