perfSONAR User Q&A and Other Discussion

[Mark Feit <>]

Petr Šesták writes:

If you would like, you can look from the last 48 hours:

 

Thanks for that.  I was able to replicate the problem in the lab and it appears to be a combination of things happening outside and inside of pScheduler.

 

The outside part is that whatever web server is being posted to is taking a very long time to complete the transaction, longer than the inside part wants.  I can’t see exactly what URL is being POSTed to because the API treats it as sensitive information and strips it out, so you’ll have to investigate that part yourself.

 

Inside of pScheduler, the plugin that does HTTP archiving doesn’t have any kind of timeout built in and will wait patiently for as long as it takes the server to respond.  The archiver service, which oversees it, has a generous two-minute timeout for a plugin to complete its business and return a result.  In your case, it looks like the plugin is taking longer than that and the software is treating that as a program startup failure.  I’ll make the error messages for that better.

 

There should be a timeout built into the HTTP archiver to prevent this from happening in the first place.  I’m always leery of imposing arbitrary timeouts, so I think the fix will be to set something reasonable (15 seconds, maybe) and add a configurable timeout to the plugin in the 5.2 release for people who know they have unusual cases.

 

In the mean time, it would be worth looking into why the destination server is taking so long to respond.

 

--Mark