Hangs when unable to connect to Loki

[erichiller]

I only recently encountered this, and it requires a specific set of circumstances it appears.
For quite a few days now my Loki server has been broken, and during all this time the Loki sink was fine, it didn't seem to have any noticeable performance impact and my program would shut down properly.

However yesterday I brought my Loki server up. Then took it down about 5 minutes later. Of course, the Loki sink worked great when the Loki server was up, but it's what happened that when it went down is interesting:
My program still ran without any noticeable performance hit, but this time required a force close to exit.

As far as I can tell this change of behaviour is because previously, the Loki server was down, but it also did not have a DNS record. After bringing it up briefly and then back down, the server had a DNS entry, but the sink was unable to contact it.
This seems to occur at CloseAndFlush().

Versions:

• Windows 10 1909
• dotnet core v3.1.7
• sink v2.2.0
• serilog v2.9.0

If there's any other info I can provide let me know.

[josephwoodward]

Oh interesting, thanks for reporting this one @erichiller, it sounds like it's related to how HttpClient works (as Loki sits over the top of the HttpSink I suspect it's using a long lived HttpClient).