Closed pierventre closed 2 years ago
Merging #482 (e294078) into main (639a6b7) will not change coverage. The diff coverage is
0.00%
.:exclamation: Current head e294078 differs from pull request most recent head 5895255. Consider uploading reports for the commit 5895255 to get more accurate results
@@ Coverage Diff @@
## main #482 +/- ##
=========================================
Coverage 69.90% 69.90%
Complexity 722 722
=========================================
Files 63 63
Lines 4748 4748
Branches 523 523
=========================================
Hits 3319 3319
Misses 1157 1157
Partials 272 272
Impacted Files | Coverage Δ | |
---|---|---|
...abric/tna/behaviour/upf/FabricUpfProgrammable.java | 60.49% <0.00%> (ø) |
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update 639a6b7...5895255. Read the comment docs.
Most secretive PR ever... ;)
Will update later the description - it is a quick and dirty fix :)
Doesn't look so ugly, do we want to merge this?
@daniele-moro are you fine to merge it after CI pass ? Or do you want to give another shot ?
@daniele-moro are you fine to merge it after CI pass ? Or do you want to give another shot ?
I'm ok in merging this after CI pass.
Sometime, we have found that UP4 commands fails in the ONOS CLI due to the following exception:
ERROR [ShellUtil] Exception caught while executing command java.lang.IllegalStateException: Context is cancelled (client has been shut down) at org.onosproject.grpc.ctl.AbstractGrpcClient.runInCancellableContext(AbstractGrpcClient.java:110) ~[?:?] at org.onosproject.p4runtime.ctl.client.P4RuntimeClientImpl.execRpc(P4RuntimeClientImpl.java:265) ~[?:?] at org.onosproject.p4runtime.ctl.client.ReadRequestImpl.submit(ReadRequestImpl.java:319) ~[?:?] at org.onosproject.p4runtime.ctl.client.ReadRequestImpl.submitSync(ReadRequestImpl.java:325) ~[?:?] at org.stratumproject.fabric.tna.behaviour.upf.FabricUpfProgrammable.readCounter(FabricUpfProgrammable.java:571)
This can easily happen as the gRPC client reference is initialized only the first time
setupBehavior
is called on any method ofFabricUpfProgrammable
. Unless the gRPC client is really broken and not available, this PR should fix the stale reference issue. Note that the skip behavior was introduced long time ago to fix a regression #289. However, this partial revert should not have any impact on the performance since the expensive part was the compute of the hw resources.