Alexey Serbin has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/12706 )

Change subject: [known_issues] the scalability of location/rack awareness
......................................................................


Patch Set 2:

(10 comments)

http://gerrit.cloudera.org:8080/#/c/12706/2/docs/known_issues.adoc
File docs/known_issues.adoc:

http://gerrit.cloudera.org:8080/#/c/12706/2/docs/known_issues.adoc@162
PS2, Line 162:    the master branch of the Kudu git repository, rebuild and 
deploy kudu-master
> Then rephrase it, e.g.
As suggested by Adar, I removed this section.


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc
File docs/known_issues.adoc:

http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@143
PS3, Line 143: rack awareness (a.k.a. location awareness)
> Why bother calling it 'rack awareness' if we call it 'location awareness' e
I thought it's necessary to mention because in the prior version of this 
document there was a mention about 'rack awareness'.  I agree it's better to 
converge on this -- removed 'rack awareness'.


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@144
PS3, Line 144: in Kudu 1.9.0 release
> Can drop this, I think.
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@145
PS3, Line 145: nnectToMaster()
             :   requests
> Can we talk about this in more abstract terms? "If the rate of new clients
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@148
PS3, Line 148: ome
> Nit: arrive
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@150
PS3, Line 150: That might manifest itself in
> That may manifest as
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@152
PS3, Line 152: alike
> so
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@151
PS3, Line 151: It's not expected to happen during regular usage of small and 
moderate-sized
             :   Kudu clusters
> I thought the main contributing factor was the number of replicas. Is that
I was not sure we wanted to go deep into the details here, but from the other 
side I agree we want go give an idea what other contributing factors are.  I 
updated this piece accordingly.

While doing testing at the scale cluster, I found that just around 10K replicas 
total and not so huge size of master's address space (under 1GB), things went 
south with the similar  request per second numbers as at the flash cluster.

I think the issue here is that the concurrent activity and size of the address 
space makes fork() to run slow.  That, in its turn, makes this pattern to grow 
as a snowball, so eventually client operations time out.


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@157
PS3, Line 157: That allows to keep the
> This retains the
Done


http://gerrit.cloudera.org:8080/#/c/12706/3/docs/known_issues.adoc@161
PS3, Line 161: ** If building from source, apply patches `ae6bbcaab` and 
`4ace91713` from
             :    the master branch of the Kudu git repository, rebuild and 
deploy kudu-master
             :    binaries.
> This doesn't seem necessary; there are always known issues with patches in
Done



--
To view, visit http://gerrit.cloudera.org:8080/12706
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: branch-1.9.x
Gerrit-MessageType: comment
Gerrit-Change-Id: I04dad488a377bf4cd36534d648a69d2fb2444fea
Gerrit-Change-Number: 12706
Gerrit-PatchSet: 2
Gerrit-Owner: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Adar Dembo <[email protected]>
Gerrit-Reviewer: Alexey Serbin <[email protected]>
Gerrit-Reviewer: Andrew Wong <[email protected]>
Gerrit-Reviewer: Greg Solovyev <[email protected]>
Gerrit-Reviewer: Kudu Jenkins (120)
Gerrit-Reviewer: Will Berkeley <[email protected]>
Gerrit-Comment-Date: Mon, 11 Mar 2019 20:11:51 +0000
Gerrit-HasComments: Yes

Reply via email to