September 2014 - Release-team - mailman3.cels.anl.gov

production deploy the MAK service
by Marcin Joachimiak 19 Sep '14

19 Sep '14

Hi All, I'm ready to do a production deploy of the MAK service. I haven't gone through a production deploy process yet, but I can successfully make deploy on my instance and use the service with remote clients. I'm attaching a list of commands that I'm using to setup the MAK service VM. I believe some parts which are a little more unique (having to do with the Java and R interaction through the JRI interface, R packages) in the end are not too bad. Mostly its recompiling R with shared library support and setting the library paths. However, I'm hoping that R recoiled for shared library support could just become part of the dev_container environment and similarly with the few environment variable settings. Let me know what you think and whether there are any changes I should be making or additional tests. best, marcin

1 0

Re: [Release-team] workspace 0.3.0 redeploy request
by Keith Keller 16 Sep '14

16 Sep '14

I talked to Gavin, and the handle manager service is a new service which must not be globally accessible, since it also needs extra privileges on shock. He says it shouldn't need much in the way of local resources. Should we add it to an existing VM? As he mentioned it also needs a Shock redeploy. How disruptive to production is that? --keith On Tue, Sep 16, 2014 at 10:49 AM, Gavin A Price <gavinaprice(a)gmail.com> wrote: > Hi all, > > I just merged v0.3.0 of the workspace service into master. Could we get a > redeploy, please? Things to note: > > - The latest version of the handle service must be running > - The latest version of the handle manager must be running (this does not > need to be globally accessible) > - a shock version => 0.8.23 must be running > - there are new entries in the config file for contacting the handle service > / manager > > Thanks, g -- Keith Keller kkeller(a)lbl.gov Physical Biosciences Division, Lawrence Berkeley National Lab

1 0

workspace 0.3.0 redeploy request
by Gavin A Price 16 Sep '14

16 Sep '14

Hi all, I just merged v0.3.0 of the workspace service into master. Could we get a redeploy, please? Things to note: - The latest version of the handle service must be running - The latest version of the handle manager must be running (this does not need to be globally accessible) - a shock version => 0.8.23 must be running - there are new entries in the config file for contacting the handle service / manager Thanks, g

1 0

Re: [Release-team] Berkeley genome annotation
by Murphy-Olson, Daniel E. 04 Sep '14

04 Sep '14

This is in place in production now. ________________________________________ From: release-team-bounces(a)lists.kbase.us [release-team-bounces(a)lists.kbase.us] on behalf of Shane Canon [scanon(a)lbl.gov] Sent: Thursday, September 04, 2014 4:54 PM To: Olson, Robert D. Cc: Keith Keller; <release-team(a)lists.kbase.us> Subject: Re: [Release-team] Berkeley genome annotation Yes. This fell of the radar to some extent. The node is ready and the service is running. It has 120GB RAM, 8 cores. It is listening at… 198.128.58.149:7050 So we just need to update nginx. —Shane On Sep 4, 2014, at 1:53 PM, Robert Olson <olson(a)mcs.anl.gov> wrote: > Shane - where are we with moving the genome annotation service to the berkeley side? > _______________________________________________ > Release-team mailing list > Release-team(a)lists.kbase.us > https://lists.kbase.us/mailman/listinfo/release-team _______________________________________________ Release-team mailing list Release-team(a)lists.kbase.us https://lists.kbase.us/mailman/listinfo/release-team

1 0

Re: [Release-team] Berkeley genome annotation
by Shane Canon 04 Sep '14

04 Sep '14

Yes. This fell of the radar to some extent. The node is ready and the service is running. It has 120GB RAM, 8 cores. It is listening at… 198.128.58.149:7050 So we just need to update nginx. —Shane On Sep 4, 2014, at 1:53 PM, Robert Olson <olson(a)mcs.anl.gov> wrote: > Shane - where are we with moving the genome annotation service to the berkeley side? > _______________________________________________ > Release-team mailing list > Release-team(a)lists.kbase.us > https://lists.kbase.us/mailman/listinfo/release-team

1 0

Berkeley genome annotation
by Robert Olson 04 Sep '14

04 Sep '14

Shane - where are we with moving the genome annotation service to the berkeley side?

1 0

Re: [Release-team] magellan
by Boisvert, Sebastien 04 Sep '14

04 Sep '14

Yes indeed. I believe that the Darwinian selection is strict here: anything (instance / volume) without any stakeholder will die in this great dying in Magellan. Also, not being able to create volume in the new OpenStack cloud makes it mostly useless. We have to wait for the old one to shut down first. Yet, the purpose of the cloud is typically to don't have to manage any of these sort of things like volume migration. In AWS, a volume never changes its name/identifier even if under the hood things can happen. ________________________________________ From: Robert Olson [olson(a)mcs.anl.gov] Sent: Thursday, September 04, 2014 3:30 PM To: Boisvert, Sebastien Cc: release-team(a)lists.kbase.us Subject: Re: [Release-team] magellan Then this model of upgrading by mass destruction is badly broken. On Sep 4, 2014, at 3:29 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: > I agree. > > But the thing is that both OpenStack deployment basically use the same storage > hardware, so there is basically not a lot of storage right now in the new cloud (that's what Scott > told me). > > > wait-as-a-service^{TM} > ________________________________________ > From: Robert Olson [olson(a)mcs.anl.gov] > Sent: Thursday, September 04, 2014 3:27 PM > To: Boisvert, Sebastien > Cc: release-team(a)lists.kbase.us > Subject: Re: [Release-team] magellan > > That seems like reason to halt any mass migration off essex. If the new volumes can't be available the new system clearly isn't ready to use. > > On Sep 4, 2014, at 3:26 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: > >> >> I think the waste of time is mostly with volumes. >> >> I migrated one instance, but it is still useless because its 400 GiB volume is not migrated yet >> by the Magellan support. >> >> ________________________________________ >> From: Robert Olson [olson(a)mcs.anl.gov] >> Sent: Thursday, September 04, 2014 3:25 PM >> To: Boisvert, Sebastien >> Cc: release-team(a)lists.kbase.us >> Subject: Re: [Release-team] magellan >> >> Thanks. What this means is that we are going to take downtime for this conversion. What a waste of time. >> >> >> On Sep 4, 2014, at 3:21 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: >> >>> You have to: >>> >>> - stop the instance cleanly in the old OpenStack ("halt" in the shell) >>> - create a snapshot of the root volume of the stopped instance >>> - any snapshop and image is synced automatically to the new OpenStack deployment >>> - in the new OpenStack cloud, boot an instance from the snapshot >>> - associate a IP address >>> - update your DNA A record, if any. >>> >>> >>> ________________________________________ >>> From: release-team-bounces(a)lists.kbase.us [release-team-bounces(a)lists.kbase.us] on behalf of Robert Olson [olson(a)mcs.anl.gov] >>> Sent: Thursday, September 04, 2014 3:13 PM >>> To: release-team(a)lists.kbase.us >>> Subject: [Release-team] magellan >>> >>> I understand that the essex instances are all getting killed tomorrow. Does anyone have experience in directly bringing across instances? I think the production genome annotation is still running there so we may lose it for a while. >>> >>> Who knows what else we may lose as well. >>> >>> >>> _______________________________________________ >>> Release-team mailing list >>> Release-team(a)lists.kbase.us >>> https://lists.kbase.us/mailman/listinfo/release-team >>> >> >> > >

1 0

Re: [Release-team] genome annotation service
by Robert Olson 04 Sep '14

04 Sep '14

No I'm not. It depends on some volumes. I need to get those migrated; we'll see what happens. On Sep 4, 2014, at 3:33 PM, Robert Olson <olson(a)mcs.anl.gov> wrote: > I am taking the genome annotation service node in a good-faith attempt to migrate since I'm utterly distracted from the productive work I was doing. I don't think we have any other notification mechanisms. > > --bob > _______________________________________________ > Release-team mailing list > Release-team(a)lists.kbase.us > https://lists.kbase.us/mailman/listinfo/release-team

1 0

genome annotation service
by Robert Olson 04 Sep '14

04 Sep '14

I am taking the genome annotation service node in a good-faith attempt to migrate since I'm utterly distracted from the productive work I was doing. I don't think we have any other notification mechanisms. --bob

1 0

Re: [Release-team] magellan
by Robert Olson 04 Sep '14

04 Sep '14

Then this model of upgrading by mass destruction is badly broken. On Sep 4, 2014, at 3:29 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: > I agree. > > But the thing is that both OpenStack deployment basically use the same storage > hardware, so there is basically not a lot of storage right now in the new cloud (that's what Scott > told me). > > > wait-as-a-service^{TM} > ________________________________________ > From: Robert Olson [olson(a)mcs.anl.gov] > Sent: Thursday, September 04, 2014 3:27 PM > To: Boisvert, Sebastien > Cc: release-team(a)lists.kbase.us > Subject: Re: [Release-team] magellan > > That seems like reason to halt any mass migration off essex. If the new volumes can't be available the new system clearly isn't ready to use. > > On Sep 4, 2014, at 3:26 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: > >> >> I think the waste of time is mostly with volumes. >> >> I migrated one instance, but it is still useless because its 400 GiB volume is not migrated yet >> by the Magellan support. >> >> ________________________________________ >> From: Robert Olson [olson(a)mcs.anl.gov] >> Sent: Thursday, September 04, 2014 3:25 PM >> To: Boisvert, Sebastien >> Cc: release-team(a)lists.kbase.us >> Subject: Re: [Release-team] magellan >> >> Thanks. What this means is that we are going to take downtime for this conversion. What a waste of time. >> >> >> On Sep 4, 2014, at 3:21 PM, Boisvert, Sebastien <boisvert(a)anl.gov> wrote: >> >>> You have to: >>> >>> - stop the instance cleanly in the old OpenStack ("halt" in the shell) >>> - create a snapshot of the root volume of the stopped instance >>> - any snapshop and image is synced automatically to the new OpenStack deployment >>> - in the new OpenStack cloud, boot an instance from the snapshot >>> - associate a IP address >>> - update your DNA A record, if any. >>> >>> >>> ________________________________________ >>> From: release-team-bounces(a)lists.kbase.us [release-team-bounces(a)lists.kbase.us] on behalf of Robert Olson [olson(a)mcs.anl.gov] >>> Sent: Thursday, September 04, 2014 3:13 PM >>> To: release-team(a)lists.kbase.us >>> Subject: [Release-team] magellan >>> >>> I understand that the essex instances are all getting killed tomorrow. Does anyone have experience in directly bringing across instances? I think the production genome annotation is still running there so we may lose it for a while. >>> >>> Who knows what else we may lose as well. >>> >>> >>> _______________________________________________ >>> Release-team mailing list >>> Release-team(a)lists.kbase.us >>> https://lists.kbase.us/mailman/listinfo/release-team >>> >> >> > >

1 0