Proget Don't retry after timeout on a connector

We have configured multiple feeds on ProGet (2 locals and one distant)
and one named "all" that simply group the 3 others in one.

We have observed multiple times that when a ProGet feed timeout on one of it's connectors, it won't retry any more to get results from that connector.

The first time it was due to an external network problem, causing our distant feed to be unavailable and our "all" feed being incomplete.
A second time it was due to an internal server maintenance that cause cause our locals feeds to not respond and once again our "all" feed was incomplete

Each time the solution was to change the connector configuration with fake changes to reboot the connector, or to reboot the ProGet IIS pool.

Having a timeout is not a real problem, especially when a real network problem has been observed.
But network problems could be "common" things, and I don't want to have to reboot ProGet each time it timeout for any reason...

Can I expect a real solution, or should I consider to recycle ProGet every hour to force itto re-evaluate connectors connectivity ?

Product: ProGet
Version: 3.2.0

We have never observed this behavior ourselves. It is already supposed to retry connections.

Just to verify, this is not something that you have observed once, and then a recycle of the app pool fixed it? It happens frequently?

It happens only 3 times, but it was 3 times in the same bad week were we had several material/network issues

But I'm sure of my assertion because I can reproduce it easily.

create a new feed
add a connector to http://nuget-prod-0-v2gallery.cloudapp.net/api/v2/ (I choose this because it's the cname equivalent of https://www.nuget.org/api/v2/ that I use for my real feed, and that I don't want to interfer with it)
try to list packages : it shoudl works
go to your host file (C:\Windows\System32\Drivers\etc\host) and add a line to re-root communication to the connector
127.0.0.1 nuget-prod-0-v2gallery.cloudapp.net
try again to list packages : it should fails (error 404, in my case, but timeout error or that is the same in the end)
remove the modification in the host file
re try and re try again, it won't works. It continuously says that there is a connector error
recycle the pool and you can successfully list package again

Thanks, we have now been able to repro this. A connector can get in this state if it times out the first time it is used after the application is initialized (There was a Lazy<> instance which kept rethrowing the initial exception on subsequent requests).

An alternative workaround to restarting the app pool is to go to the edit feed page for the stuck feed, and click Save. That should force cached values to be discarded.

We have also fixed this in the next build; the fix will be included in 3.2.1.

Thanks again for the detailed repro steps!

davidroberts63

I'm still seeing that caching problem that you mentioned.
v3.8.1(build10) IIS hosted

Edit feed that has no connector
Add connector "http://localhost" (for testing, others can do the same as long as it errors on first try)
View the feed, see the delayed 'loading' and then the 'error with a connector' at top
Edit feed again. Remove that connector (just the red x)
View the feed, see the delayed 'loading' and then the 'error with a connector' at top
Edit feed again, change description (or something to trigger a save).
View the feed, see the delayed 'loading' and then the 'error with a connector' at top
Recycle app pool
View the feed. Problem gone.

I repeated the above process three times to make sure. I also waited about five minutes between step 4 and step 5 to see if that would do anything.

Thanks for the report, I logged this as an issue to further investigate... PG-492