Repository navigation
DNS tests are failing deterministically in CI #48751
Description
System.Net.Internals.SocketExceptionFactory+ExtendedSocketException : Name or service not known
at System.Net.Dns.GetHostEntryOrAddressesCore(String hostName, Boolean justAddresses, AddressFamily addressFamily) in /_/src/libraries/System.Net.NameResolution/src/System/Net/Dns.cs:line 404
at System.Net.Dns.GetHostEntryCore(String hostName, AddressFamily addressFamily) in /_/src/libraries/System.Net.NameResolution/src/System/Net/Dns.cs:line 385
at System.Net.Dns.GetHostByName(String hostName) in /_/src/libraries/System.Net.NameResolution/src/System/Net/Dns.cs:line 272
at System.Net.NameResolution.Tests.GetHostByNameTest.DnsObsoleteGetHostByName_EmptyString_ReturnsHostName() in /_/src/libraries/System.Net.NameResolution/tests/FunctionalTests/GetHostByNameTest.cs:line 108
We probably need to disable the failing tests right now as apparently something changed in the dns detection which is causing these tests to fail deterministically in CI.
cc @dotnet/ncl
Activity
wfurt commentedon Feb 25, 2021
Do you know if anything changed with infrastructure @MattGal ? From the logs it seems like the instance cannot resolve it's own name. I'll keep looking as it is not clear why this would happen only on particular distro.
MattGal commentedon Feb 25, 2021
Interesting; I know of no intentional changes from this but Weds. is rollout days so if something changed, it would have been yesterday. Taking a look.
MattGal commentedon Feb 25, 2021
Ah. SLES 15. Yes, we did update this one yesterday, because SUSE deleted the SLES 15 SP1 image from the gallery, forcing us onto SP2. I'll check it out and see if there's anything we can do and report back. Tracking issue: https://github.com/dotnet/core-eng/issues/12333
wfurt commentedon Feb 25, 2021
thanks @MattGal. If needed, I can help with the investigation.
wfurt commentedon Feb 25, 2021
cc @dnceng for visibility.
MattGal commentedon Feb 25, 2021
@dotnet/dnceng also
MattGal commentedon Feb 25, 2021
the problem comes from including git-core, I am looking at working around it.
VincentBu commentedon Feb 26, 2021
Failed again in runtime 20210225.61
Failed test:
Error message:
1 remaining item
MattGal commentedon Feb 26, 2021
@VincentBu yes, while I have a lead on what caused this, the earliest this will be fixed on the machines is next Weds. I was under the impression from @wfurt this test was disabled on that platform for the time being?
ViktorHofer commentedon Feb 26, 2021
Yes, the tests are already disabled.
karelz commentedon Apr 23, 2021
@MattGal @ViktorHofer did the DNS get fixed in CI? Should we re-enable the tests?
MattGal commentedon Apr 23, 2021
It's hard to be certain since the tests are off, but in retrospect we were definitely having global issues with DNS in Azure machines that have subsided (down to a single domain that may have problems from creating hosts fast?) in the March timeframe.
I'd say it's worth a try, since various things have changed since Mar. 3. I'll take a stab when I finish my inbox but the tricky part is with your test disabled I don't have a simple repro path.
MattGal commentedon May 4, 2021
@karelz sorry for the delay. The workaround I made for the strange symptom from before (inability to resolve the local host) is still holding, so yes you should reenable these tests on SLES 15.
Re-enable DNS tests on SLES, which were failing due to infra - see do…
Re-enable DNS tests on SLES