Test suite overhaul parallel #2333
Conversation
Each of these tests relies on a write of only a few points. It simply should not take 60 seconds.
| @@ -25,6 +25,9 @@ const ( | |||
| // Use a prime batch size, so that internal batching code, which most likely | |||
| // uses nice round batches, has to deal with leftover. | |||
| batchSize = 4217 | |||
|
|
|||
| openTSDBTestTimeout = 10 * time.Second | |||
There was a problem hiding this comment.
Do these really take this long sometimes?
There was a problem hiding this comment.
Probably not since, I finally fixed the inputs. Easy to change.
|
Some question around the listen/accept with return vs. continue. Otherwise, looks awesome! Super excited to see this in. Thanks for all the hard work! |
|
👍 from me after cory's comments |
|
openTSDB and Graphite TCP listeners only return when a specific error is returned. |
|
| for { | ||
| conn, err := ln.Accept() | ||
| if _, ok := err.(*net.OpError); ok && strings.Contains(err.Error(), "use of closed network connection") { |
There was a problem hiding this comment.
I tested this. This is the error that comes out when the listener is closed.
|
+1 |
|
@otoolep Just did some more research and the right way to handle that error is to see if it's a TempError: http://golang.org/pkg/net/#OpError.Temporary If it's not, then it's for sure game over. |
|
I'll revert logging heartbeat failures at trace. I have a different way to address the issue. |
Other types of Accept() failures should result in a retry.
otoolep
force-pushed
the
test_suite_overhaul_parallel
branch
from
0ffa9ee to
0b1fc43
Compare
|
All comments addressed. 17 test runs, 1 failure (clustering issue?) 16 passes. Merging. |
| if h, _ = os.Hostname(); h == "" { | ||
| h = "localhost" | ||
| } | ||
| } |
There was a problem hiding this comment.
on OSX, if i run influxd -hostname localhost, h will initially be "::" and then will become the result of os.Hostname() when what i really want is for it to be "localhost" (or whatever i specify via -hostname). it was working that way before this change.
- any ideas on how to make it work again?
- thoughts on how to add a test to prevent a regression?
- ...and why do you treat
"::"differently to"0.0.0.0"? shouldn't they either both be changed or not changed?
tracked at #2350
No description provided.