connection refuseLeap status not available

2017-05-24 Thread Jim Apple
testdata/cluster/admin

calls ntp-wait which returns an error:

"ntpq: read: Connection refuseLeap status not avalaible"

Has anyone seen this? There are 0 Google hits for refuseLeap and the
string is not present in my repository or my /etc/.


Re: connection refuseLeap status not available

2017-05-24 Thread Tim Armstrong
I see that with some frequency when restarting my system.

I usually manage to fix it with a non-scientific approach of running some
combination of these commands until ntp-wait works:

sudo service ntp restart
sudo ntpdate -s ntp.ubuntu.com
ntp-wait -v

On Wed, May 24, 2017 at 11:33 AM, Jim Apple  wrote:

> testdata/cluster/admin
>
> calls ntp-wait which returns an error:
>
> "ntpq: read: Connection refuseLeap status not avalaible"
>
> Has anyone seen this? There are 0 Google hits for refuseLeap and the
> string is not present in my repository or my /etc/.
>


Re: connection refuseLeap status not available

2017-05-24 Thread Zachary Amsden
This is similar to what I hit with NTP the other day after a restart.  I
tried a number of things, and I think the only thing that worked was
waiting for NTP to sync.  Pitfalls: ntpdate requires a host on the command
line, and doesn't read the configuration file.

There was some circumstantial evidence that a connection to the Ubuntu NTP
pool failed during boot - overload or random connection failure, but I
never got an exact cause.

On Wed, May 24, 2017 at 12:09 PM, Tim Armstrong 
wrote:

> I see that with some frequency when restarting my system.
>
> I usually manage to fix it with a non-scientific approach of running some
> combination of these commands until ntp-wait works:
>
> sudo service ntp restart
> sudo ntpdate -s ntp.ubuntu.com
> ntp-wait -v
>
> On Wed, May 24, 2017 at 11:33 AM, Jim Apple  wrote:
>
> > testdata/cluster/admin
> >
> > calls ntp-wait which returns an error:
> >
> > "ntpq: read: Connection refuseLeap status not avalaible"
> >
> > Has anyone seen this? There are 0 Google hits for refuseLeap and the
> > string is not present in my repository or my /etc/.
> >
>


Re: connection refuseLeap status not available

2017-05-24 Thread Tim Armstrong
Yes it's also possible that my solution "works" because it distracts me for
long enough for NTP to sync :)

On Wed, May 24, 2017 at 12:20 PM, Zachary Amsden 
wrote:

> This is similar to what I hit with NTP the other day after a restart.  I
> tried a number of things, and I think the only thing that worked was
> waiting for NTP to sync.  Pitfalls: ntpdate requires a host on the command
> line, and doesn't read the configuration file.
>
> There was some circumstantial evidence that a connection to the Ubuntu NTP
> pool failed during boot - overload or random connection failure, but I
> never got an exact cause.
>
> On Wed, May 24, 2017 at 12:09 PM, Tim Armstrong 
> wrote:
>
> > I see that with some frequency when restarting my system.
> >
> > I usually manage to fix it with a non-scientific approach of running some
> > combination of these commands until ntp-wait works:
> >
> > sudo service ntp restart
> > sudo ntpdate -s ntp.ubuntu.com
> > ntp-wait -v
> >
> > On Wed, May 24, 2017 at 11:33 AM, Jim Apple 
> wrote:
> >
> > > testdata/cluster/admin
> > >
> > > calls ntp-wait which returns an error:
> > >
> > > "ntpq: read: Connection refuseLeap status not avalaible"
> > >
> > > Has anyone seen this? There are 0 Google hits for refuseLeap and the
> > > string is not present in my repository or my /etc/.
> > >
> >
>


Re: connection refuseLeap status not available

2017-05-24 Thread Jim Apple
'sudo service ntp restart' WFM. Thanks!

On Wed, May 24, 2017 at 12:09 PM, Tim Armstrong  wrote:
> I see that with some frequency when restarting my system.
>
> I usually manage to fix it with a non-scientific approach of running some
> combination of these commands until ntp-wait works:
>
> sudo service ntp restart
> sudo ntpdate -s ntp.ubuntu.com
> ntp-wait -v
>
> On Wed, May 24, 2017 at 11:33 AM, Jim Apple  wrote:
>
>> testdata/cluster/admin
>>
>> calls ntp-wait which returns an error:
>>
>> "ntpq: read: Connection refuseLeap status not avalaible"
>>
>> Has anyone seen this? There are 0 Google hits for refuseLeap and the
>> string is not present in my repository or my /etc/.
>>


GVO machines dependencies

2017-05-24 Thread Sailesh Mukil
I have a question regarding dependencies on the jenkins machines. (
http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/)

How do the Jenkins machines used to GVO have dependencies pre-packaged in
them? For eg., openssl, etc. Are they baked into an AMI?

I have a patch (https://gerrit.cloudera.org/#/c/6910/) that requires a new
dependency for Ubuntu (libffi), and therefore won't pass GVO. This didn't
show up in all my testing since the machines I used already had this
dependency in them.

Is there a way to add a new dependency to these machines?


Re: GVO machines dependencies

2017-05-24 Thread Jim Apple
They run https://github.com/awleblang/impala-setup. One of the jobs
run as part of this is
http://jenkins.impala.io:8080/view/Utility/job/ubuntu-14.04-build-only/
which runs ./bin/bootstrap_build.sh (in the Apache Impala repo), so
you'll need to update that file, too, I suspect.



On Wed, May 24, 2017 at 2:56 PM, Sailesh Mukil  wrote:
> I have a question regarding dependencies on the jenkins machines. (
> http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/)
>
> How do the Jenkins machines used to GVO have dependencies pre-packaged in
> them? For eg., openssl, etc. Are they baked into an AMI?
>
> I have a patch (https://gerrit.cloudera.org/#/c/6910/) that requires a new
> dependency for Ubuntu (libffi), and therefore won't pass GVO. This didn't
> show up in all my testing since the machines I used already had this
> dependency in them.
>
> Is there a way to add a new dependency to these machines?


Re: GVO machines dependencies

2017-05-24 Thread Michael Brown
> Is there a way to add a new dependency to these machines?

Our builds use two worker labels:
1. ubuntu14.04-c4.4xlarge-gp2
2. ub14-build-only

1 uses https://github.com/awleblang/impala-setup, and I think you should
add libffi to there. It's something you should do anyway so new users get
set up with the right dependencies.

2 is a different story, but I see that at
http://jenkins.impala.io:8080/configure that worker is already installing a
JDK via apt-get. You could use a similar pattern to install libffi.


Re: GVO machines dependencies

2017-05-24 Thread Jim Apple
I'd recommend against #2 using the node setup to install Impala
dependencies. The only reason it installs openjdk is that Jenkins
needs Java to talk to it. All of the other Impala dependencies are
installed in the jobs themselves.

On Wed, May 24, 2017 at 3:11 PM, Michael Brown  wrote:
>> Is there a way to add a new dependency to these machines?
>
> Our builds use two worker labels:
> 1. ubuntu14.04-c4.4xlarge-gp2
> 2. ub14-build-only
>
> 1 uses https://github.com/awleblang/impala-setup, and I think you should
> add libffi to there. It's something you should do anyway so new users get
> set up with the right dependencies.
>
> 2 is a different story, but I see that at
> http://jenkins.impala.io:8080/configure that worker is already installing a
> JDK via apt-get. You could use a similar pattern to install libffi.


Re: GVO machines dependencies

2017-05-24 Thread Michael Brown
Thanks Jim; I had forgotten bin/bootstrap_build.sh.

On Wed, May 24, 2017 at 3:14 PM, Jim Apple  wrote:

> I'd recommend against #2 using the node setup to install Impala
> dependencies. The only reason it installs openjdk is that Jenkins
> needs Java to talk to it. All of the other Impala dependencies are
> installed in the jobs themselves.
>
> On Wed, May 24, 2017 at 3:11 PM, Michael Brown  wrote:
> >> Is there a way to add a new dependency to these machines?
> >
> > Our builds use two worker labels:
> > 1. ubuntu14.04-c4.4xlarge-gp2
> > 2. ub14-build-only
> >
> > 1 uses https://github.com/awleblang/impala-setup, and I think you should
> > add libffi to there. It's something you should do anyway so new users get
> > set up with the right dependencies.
> >
> > 2 is a different story, but I see that at
> > http://jenkins.impala.io:8080/configure that worker is already
> installing a
> > JDK via apt-get. You could use a similar pattern to install libffi.
>


Re: GVO machines dependencies

2017-05-24 Thread Sailesh Mukil
Thanks Michael and Jim,

It looks like adding to bin/bootstrap_build.sh makes the ub14-build-only
job work fine, however, updating the impala-setup still doesn't seem to
work. I submitted a pull request that was merged by Dimitris here:
https://github.com/awleblang/impala-setup/commits/master

Also, the impala-setup chef logs don't seem to get printed in the jenkins
console output. So I'm not sure if it's actually being run or not.

On Wed, May 24, 2017 at 3:37 PM, Michael Brown  wrote:

> Thanks Jim; I had forgotten bin/bootstrap_build.sh.
>
> On Wed, May 24, 2017 at 3:14 PM, Jim Apple  wrote:
>
> > I'd recommend against #2 using the node setup to install Impala
> > dependencies. The only reason it installs openjdk is that Jenkins
> > needs Java to talk to it. All of the other Impala dependencies are
> > installed in the jobs themselves.
> >
> > On Wed, May 24, 2017 at 3:11 PM, Michael Brown 
> wrote:
> > >> Is there a way to add a new dependency to these machines?
> > >
> > > Our builds use two worker labels:
> > > 1. ubuntu14.04-c4.4xlarge-gp2
> > > 2. ub14-build-only
> > >
> > > 1 uses https://github.com/awleblang/impala-setup, and I think you
> should
> > > add libffi to there. It's something you should do anyway so new users
> get
> > > set up with the right dependencies.
> > >
> > > 2 is a different story, but I see that at
> > > http://jenkins.impala.io:8080/configure that worker is already
> > installing a
> > > JDK via apt-get. You could use a similar pattern to install libffi.
> >
>


Re: GVO machines dependencies

2017-05-24 Thread Michael Brown
"from-scratch" is a misnomer: only the first build to run on a given node
is truly from-scratch, because running impala-setup takes effect
system-wide, and is run once, when the node is created.

For http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/1371/ , the
worker's build history suggests it has been up roughly 17 hours.
http://jenkins.impala.io:8080/computer/ub1404-c4.4xl-gp2%20(i-0d76efbc8de26926c)/builds
This means your recent change hasn't taken effect on "older" workers.

Any new workers that get created should get the change.





On Wed, May 24, 2017 at 4:34 PM, Sailesh Mukil  wrote:

> Thanks Michael and Jim,
>
> It looks like adding to bin/bootstrap_build.sh makes the ub14-build-only
> job work fine, however, updating the impala-setup still doesn't seem to
> work. I submitted a pull request that was merged by Dimitris here:
> https://github.com/awleblang/impala-setup/commits/master
>
> Also, the impala-setup chef logs don't seem to get printed in the jenkins
> console output. So I'm not sure if it's actually being run or not.
>
> On Wed, May 24, 2017 at 3:37 PM, Michael Brown  wrote:
>
> > Thanks Jim; I had forgotten bin/bootstrap_build.sh.
> >
> > On Wed, May 24, 2017 at 3:14 PM, Jim Apple  wrote:
> >
> > > I'd recommend against #2 using the node setup to install Impala
> > > dependencies. The only reason it installs openjdk is that Jenkins
> > > needs Java to talk to it. All of the other Impala dependencies are
> > > installed in the jobs themselves.
> > >
> > > On Wed, May 24, 2017 at 3:11 PM, Michael Brown 
> > wrote:
> > > >> Is there a way to add a new dependency to these machines?
> > > >
> > > > Our builds use two worker labels:
> > > > 1. ubuntu14.04-c4.4xlarge-gp2
> > > > 2. ub14-build-only
> > > >
> > > > 1 uses https://github.com/awleblang/impala-setup, and I think you
> > should
> > > > add libffi to there. It's something you should do anyway so new users
> > get
> > > > set up with the right dependencies.
> > > >
> > > > 2 is a different story, but I see that at
> > > > http://jenkins.impala.io:8080/configure that worker is already
> > > installing a
> > > > JDK via apt-get. You could use a similar pattern to install libffi.
> > >
> >
>


Re: GVO machines dependencies

2017-05-24 Thread Sailesh Mukil
The issue was that there was a bug in the install.sh script from the
impala-setup/ repo that always assumed that the repo would reside in
/home/$user.

The Jenkins AMIs seem to have had a very old checkout (8 months old) at
/home/ubuntu/ which was being used for the GVOs. The Jenkins job would make
a fresh checkout in /tmp/ and run install.sh from that checkout. But due to
the above bug, the script would change its sources to the old repo in
/home/ubuntu. So, the new dependencies added to the project didn't get
picked up. I've submitted a fix, and will redo the GVO once it's merged
into impala-setup.

Thanks for your help Michael and Jim!

On Wed, May 24, 2017 at 4:58 PM, Michael Brown  wrote:

> "from-scratch" is a misnomer: only the first build to run on a given node
> is truly from-scratch, because running impala-setup takes effect
> system-wide, and is run once, when the node is created.
>
> For http://jenkins.impala.io:8080/job/ubuntu-14.04-from-scratch/1371/ ,
> the
> worker's build history suggests it has been up roughly 17 hours.
> http://jenkins.impala.io:8080/computer/ub1404-c4.4xl-gp2%20(
> i-0d76efbc8de26926c)/builds
> This means your recent change hasn't taken effect on "older" workers.
>
> Any new workers that get created should get the change.
>
>
>
>
>
> On Wed, May 24, 2017 at 4:34 PM, Sailesh Mukil 
> wrote:
>
> > Thanks Michael and Jim,
> >
> > It looks like adding to bin/bootstrap_build.sh makes the ub14-build-only
> > job work fine, however, updating the impala-setup still doesn't seem to
> > work. I submitted a pull request that was merged by Dimitris here:
> > https://github.com/awleblang/impala-setup/commits/master
> >
> > Also, the impala-setup chef logs don't seem to get printed in the jenkins
> > console output. So I'm not sure if it's actually being run or not.
> >
> > On Wed, May 24, 2017 at 3:37 PM, Michael Brown 
> wrote:
> >
> > > Thanks Jim; I had forgotten bin/bootstrap_build.sh.
> > >
> > > On Wed, May 24, 2017 at 3:14 PM, Jim Apple 
> wrote:
> > >
> > > > I'd recommend against #2 using the node setup to install Impala
> > > > dependencies. The only reason it installs openjdk is that Jenkins
> > > > needs Java to talk to it. All of the other Impala dependencies are
> > > > installed in the jobs themselves.
> > > >
> > > > On Wed, May 24, 2017 at 3:11 PM, Michael Brown 
> > > wrote:
> > > > >> Is there a way to add a new dependency to these machines?
> > > > >
> > > > > Our builds use two worker labels:
> > > > > 1. ubuntu14.04-c4.4xlarge-gp2
> > > > > 2. ub14-build-only
> > > > >
> > > > > 1 uses https://github.com/awleblang/impala-setup, and I think you
> > > should
> > > > > add libffi to there. It's something you should do anyway so new
> users
> > > get
> > > > > set up with the right dependencies.
> > > > >
> > > > > 2 is a different story, but I see that at
> > > > > http://jenkins.impala.io:8080/configure that worker is already
> > > > installing a
> > > > > JDK via apt-get. You could use a similar pattern to install libffi.
> > > >
> > >
> >
>