Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
Created a quick attempt at a static parent span to pass to the threads in
7444888ff8582bc31d0c61444103cc4dc9ff362a
Doesn't seem to work though whomp whomp
---
If your project is
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
> If we can create spans by explicitly passing the parent (constructor or
static method)
I'm not quite sure exactly what you mean... instead of this:
`Span span =
Github user ctubbsii commented on the issue:
https://github.com/apache/accumulo/pull/255
The reason this is borked due to the thread pools is probably due to the
fact that the parent span id is stored in a thread local variable. If we can
create spans by explicitly passing the parent
Github user ctubbsii commented on the issue:
https://github.com/apache/accumulo/pull/255
Also, GitHub appears broke right now... so if you see the same comment
repeated 4x sometime later, it's because I tried to submit it as a review
comment a few times first, but it didn't work.
Github user ctubbsii commented on the issue:
https://github.com/apache/accumulo/pull/255
Instead of long javadoc, could just leave old behavior in, commented-out
with comment line above it explaining that the span doesn't always appear
(isn't always rooted properly).
---
If your
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/255
> Could it be because we have the traces in 2 separate thread pools and
htrace gets confused:
> Or because we change the name of the thread (wtf??) in
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
Or because we change the name of the thread (wtf??) in
TabletServerBatchWriter.SendTask.send()
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
Could it be because we have the traces in 2 separate thread pools and
htrace gets confused:
In org/apache/accumulo/core/client/impl/TabletServerBatchWriter.java
BinMutations:
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/255
> Now why is it rooted correctly when run manually in the shell
So, the hierarchy definitely seems goofed to me.
MutationWriter+sendMutations are contained beneath BinMutations (i'd
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
Now why is it rooted correctly when run manually in the shell
sh...@canx2um-78286w.infosec.tycho.ncsc.mil close
46+8015 sh...@canx2um-78286w.infosec.tycho.ncsc.mil
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/255
Aha!
> The following spans are not rooted (probably due to a parent span of
length 0ms):
I don't think this test is even validating the correct thing. That span is
supposed to
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
FAILURE running with Maven:
Trace started at 2017/04/28 11:06:53.041
Time Start Service@Location Name
2251+0 sh...@canx2um-78286w.infosec.tycho.ncsc.mil
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
SUCCESS running with Maven:
Trace started at 2017/04/28 11:01:58.382
Time Start Service@Location Name
2375+0 sh...@canx2um-78286w.infosec.tycho.ncsc.mil
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
Shell trace output:
Trace started at 2017/04/28 10:49:35.021
Time Start Service@Location Name
25863+0 sh...@canx2um-78286w.infosec.tycho.ncsc.mil shell:root
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/255
> I don't think sendMutations get lost, just that its a race condition we
don't have control over.
Correct me if I'm wrong: but if we just don't understand the problem, I'm
not in
Github user adamjshook commented on the issue:
https://github.com/apache/accumulo/pull/254
Thanks for the pointers. I'll head down this route and get back to you
soon.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If
Github user milleruntime commented on the issue:
https://github.com/apache/accumulo/pull/255
> Do you have more specifics as to why this is?
No, only experience with getting different traces and your and Shawn
Walkers comments on
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/254
> the replicate call takes in the ReplicaSystemHelper as an argument and I
wasn't too sure how to go about creating an instance of that
Nothing special, just instantiate it. It requires
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/255
> The multithreaded nature of mutations are now too unpredictable and time
sensitive to reliably look for "sendMutations" traces here.
Do you have more specifics as to why this is?
Github user adamjshook commented on the issue:
https://github.com/apache/accumulo/pull/254
@joshelser I think so. I was taking a look at `AccumuloReplicaSystemTest`
and seeing how I could fit something in to try and trigger the timeout, but the
`replicate` call takes in the
GitHub user milleruntime opened a pull request:
https://github.com/apache/accumulo/pull/255
ACCUMULO-4365: Fix trace test in ShellServerIT
This test has been frustrating for some time now... After spending some
time looking into a potential bug (possibly one similar to
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/244
> I understand that when you take into consideration all the other factors
of a cluster, low level system iterators only make up a small percentage of
system performance considerations
Github user joshelser commented on the issue:
https://github.com/apache/accumulo/pull/254
Thanks, @adamjshook.
> Not too sure how to go about adding a unit or integration test for this
timeout
Would it help to re-factor this code to make it more unit-testable? I
Github user adamjshook commented on the issue:
https://github.com/apache/accumulo/pull/254
@joshelser Applied your feedback to the commit. I'll have to run it
through my test later, though -- away from my PC at the moment.
---
If your project is set up for it, you can reply to this
For deleting test tables, you may find the flag "-p" for the deletetable
shell command useful. For example,
deletetable -f -p Test.*
will delete all tables that begin with the prefix "Test". You could even
call this from a script if you want to automate it into your build.
Perhaps this would
25 matches
Mail list logo
