cmd/workload: implement checks for history-pruned node #31355
Conversation
|
It's not supposed to skip these tests, it should still call the function but check that the returned error has the correct error code (4444). |
|
Yesterday, you said that this could be worked on right now even though we don't currently have the history-truncation rpc error return implemented. I took that to mean that we would just skip affected test cases. |
|
Yeah, so what I mean is, these tests would fail against a pruned node. But that's OK. They already fail now because the node returns an arbitrary error. |
|
Okay, so afaict this is blocked until we can surface the pruned history error from the relevant rpc calls, which itself requires the dependent components under the hood to properly distinguish it. |
|
No it's not. You can just add a verification for the error code here. |
jwasinger
force-pushed
the
workload-skip-premerge
branch
from
aa090ca to
9d5c9f3
Compare
|
Maybe add a helper function for this check |
|
Yeah agreed, it's a bit verbose. |
cmd/workload/historytest.go
Outdated
| // code of 4444. | ||
| func errIsPrunedHistory(err error) bool { | ||
| if err != nil { | ||
| if rpcErr := err.(rpc.Error); rpcErr != nil && rpcErr.ErrorCode() == 4444 { |
There was a problem hiding this comment.
rpcErr := err.(rpc.Error) will crash if it doesn't satisfy the interface. You need to capture the second return value to check if it was OK.
|
TODO (self): ensure in the tests that if the 4444 code is emitted when the test block (or range) is passed the prune threshold, the tests error. |
fjl
changed the title
…g failure to access missing pruned history, don't mark that test run as a failure
…om pruned history
…ning threshold. if so, it is an error
|
lint is red, also pls add a pr description |
fjl
force-pushed
the
workload-skip-premerge
branch
from
3bb3667 to
b0eaf5f
Compare
Co-authored-by: Felix Lange <[email protected]>
Co-authored-by: Felix Lange <[email protected]>
Co-authored-by: Felix Lange <[email protected]>
No description provided.