We seem to run across this error at seemingly random intervals on our server services. The full text of the error is:
Error: Failure to access the DBMS server [08S01:[Microsoft][SQL Server Native Client 11.0]TCP Provider: An established connection was aborted by the software in your host machine. ]
Reading up on Azure, it looks like Azure is closing the db connections on their own for...reasons? Anyway, is there a solution to this? So far the only thing I've found that works is either restarting the service or waiting for the server to shut down an idle instance and restart a new one. I've even tried putting a small python script on the scheduler that will hit the service with a random query every 10 or 15 minutes to no avail.
Specs:
Arcgis Server 10.3 Enterprise Standard (vanilla install..no extensions, no portal)
Azure A3 VM running 2012 R2 Enterprise Server (4 cores, 7 gigs ram)
Azure S2 Database (50 DTUs)
Our database is ~3.5 gigs with around a million records or so.
Thanks. The Azure platform with server is awesome otherwise.
Hi Todd,
I recommend that you use a higher performance Tier for Azure SQL Database.
I would suggest at-least an S2 or higher for anything other than a development environment.
Thanks
-Nik
Thanks for the replyl Nik. We're using the S2 level for our stuff right now but I guess I could bump that up to S3 if that's really the answer. I'm concerned, thought, that after the S3 tier, we'd have to jump to the "P" level tier which is really out of range for us cost wise.
From what I've read, it seems like Azure is purposely dropping connections based on overall load of the shared resources and that the client software (AGS Server in this case) is supposed to handle this somehow and reconnect? I'm just not sure how to make this happen short of restarting the services.
I'm headed out to the Dev Summit in a few weeks so perhaps I can pester you guys some more about this out there? LOL
Hi Todd,
Your observations are correct about SQL Azure and Feature Services.
We can discuss at Dev Summit about potential workarounds - service restarts are one possibility.
Did you ever come up with a solution on this? If so, could you post it?
Thanks greatly.
I know this is a late reply, but no we never did figure out a solution. In the end we were unfortunately forced to migrate to Amazon which wasn't terribly (in our case at least).