Hi everyone,
As you probably read from CT's email last week about the Eqiad migration (you read that, right? No? Go read it. I'll wait)
Ok, done now? So, as you know (now), we plan to suspend all scheduled deployments next week during the data center migration (January 21-25). We'd like to keep the number of moving parts to a minimum, thus we'd like to avoid changing anything unless necessary.
That said, we are contemplating leaving the Wednesday window for 1.21wmf8 to non-Wikipedia in place as a means of testing deployments themselves. Unlike previous core deployments (which are pretty routine), we're reserving the right not to have that deployment, and may delay the already-revised deployment calendar one more week if things are still messy on Wednesday. I'm guessing that the odds of that deployment happen are pretty low (I'll say "20%" based on aerial extraction).
So, to reiterate the plan with respect to deployments over the next couple of weeks:
January 14-15 (Mon-Tue): deployments from fenari to Tampa cluster using scap/sync-file/sync-dir as normal. January 16 (Wed): 1.21wmf8 deployed to test, test2, and mediawiki.org January 16-18 (Wed-Fri): 1.21wmf8 deployments using git-deploy from fenari January 17 (Thu): Brown bag with Ryan Lane and Chris Steipp to talk about git-deploy January 21 (Mon): Deployment freeze (and U.S. holiday) January 22-25 (Tue-Fri): Deployment freeze January 23 (Wed): MAYBE 1.21wmf8 deployment to some or all non-Wikipedia sites.
I hope this all makes sense. Questions/comments/concerns?
Rob
On the date you plan to deploy 1.21wmf8 to test, test2, and mediawiki; do you also plan on deploying it to Wikidata with them or with the other non-Wikipedias?
Techman224
On 2013-01-14, at 7:26 PM, Rob Lanphier robla@wikimedia.org wrote:
Hi everyone,
As you probably read from CT's email last week about the Eqiad migration (you read that, right? No? Go read it. I'll wait)
Ok, done now? So, as you know (now), we plan to suspend all scheduled deployments next week during the data center migration (January 21-25). We'd like to keep the number of moving parts to a minimum, thus we'd like to avoid changing anything unless necessary.
That said, we are contemplating leaving the Wednesday window for 1.21wmf8 to non-Wikipedia in place as a means of testing deployments themselves. Unlike previous core deployments (which are pretty routine), we're reserving the right not to have that deployment, and may delay the already-revised deployment calendar one more week if things are still messy on Wednesday. I'm guessing that the odds of that deployment happen are pretty low (I'll say "20%" based on aerial extraction).
So, to reiterate the plan with respect to deployments over the next couple of weeks:
January 14-15 (Mon-Tue): deployments from fenari to Tampa cluster using scap/sync-file/sync-dir as normal. January 16 (Wed): 1.21wmf8 deployed to test, test2, and mediawiki.org January 16-18 (Wed-Fri): 1.21wmf8 deployments using git-deploy from fenari January 17 (Thu): Brown bag with Ryan Lane and Chris Steipp to talk about git-deploy January 21 (Mon): Deployment freeze (and U.S. holiday) January 22-25 (Tue-Fri): Deployment freeze January 23 (Wed): MAYBE 1.21wmf8 deployment to some or all non-Wikipedia sites.
I hope this all makes sense. Questions/comments/concerns?
Rob
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
I am scheduled to do a deployment (to all wikis) for MobileFrontend this Thursday afternoon: * I presume scap/sync-blah will still work for 1.21wmf7 - can you please confirm? * Will I *have* to use git-deploy for 1.21wmf8 or will scap/sync-blah still work?
And to reiterate my question from a few days ago re the deployment freeze next week: * During the 'deployment freeze', in the event that someone needs to deploy an emergency fix, what do we do and who do we need to communicate with? Will scap/deploy-whatever still be functional, or will gti-deploy be the de-facto deployment method at that point?
On Mon, Jan 14, 2013 at 6:51 PM, Techman224 techman224@techman224.cawrote:
On the date you plan to deploy 1.21wmf8 to test, test2, and mediawiki; do you also plan on deploying it to Wikidata with them or with the other non-Wikipedias?
Techman224
On 2013-01-14, at 7:26 PM, Rob Lanphier robla@wikimedia.org wrote:
Hi everyone,
As you probably read from CT's email last week about the Eqiad migration (you read that, right? No? Go read it. I'll wait)
Ok, done now? So, as you know (now), we plan to suspend all scheduled deployments next week during the data center migration (January 21-25). We'd like to keep the number of moving parts to a minimum, thus we'd like to avoid changing anything unless necessary.
That said, we are contemplating leaving the Wednesday window for 1.21wmf8 to non-Wikipedia in place as a means of testing deployments themselves. Unlike previous core deployments (which are pretty routine), we're reserving the right not to have that deployment, and may delay the already-revised deployment calendar one more week if things are still messy on Wednesday. I'm guessing that the odds of that deployment happen are pretty low (I'll say "20%" based on aerial extraction).
So, to reiterate the plan with respect to deployments over the next couple of weeks:
January 14-15 (Mon-Tue): deployments from fenari to Tampa cluster using scap/sync-file/sync-dir as normal. January 16 (Wed): 1.21wmf8 deployed to test, test2, and mediawiki.org January 16-18 (Wed-Fri): 1.21wmf8 deployments using git-deploy from
fenari
January 17 (Thu): Brown bag with Ryan Lane and Chris Steipp to talk about git-deploy January 21 (Mon): Deployment freeze (and U.S. holiday) January 22-25 (Tue-Fri): Deployment freeze January 23 (Wed): MAYBE 1.21wmf8 deployment to some or all non-Wikipedia
sites.
I hope this all makes sense. Questions/comments/concerns?
Rob
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
Hi Arthur,
Sorry for the delayed reply. Comments below:
On Tue, Jan 15, 2013 at 12:07 PM, Arthur Richards arichards@wikimedia.org wrote:
I am scheduled to do a deployment (to all wikis) for MobileFrontend this Thursday afternoon:
- I presume scap/sync-blah will still work for 1.21wmf7 - can you please
confirm?
After tomorrow's deployment, assuming it goes well, it's going to be all git-deploy, all the time, and scap/sync-file will go away.
- Will I *have* to use git-deploy for 1.21wmf8 or will scap/sync-blah still
work?
You'll need to use git-deploy.
And to reiterate my question from a few days ago re the deployment freeze next week:
- During the 'deployment freeze', in the event that someone needs to deploy
an emergency fix, what do we do and who do we need to communicate with?
Use #wikimedia-operations. If no one is around, and it is an emergency, go ahead and make the fix, and note it in the logs.
Will scap/deploy-whatever still be functional, or will gti-deploy be the de-facto deployment method at that point?
git-deploy is going to be it.
Rob
Thanks for the response, Rob (and I hope you're feeling better soon). I'm terrified about using new tools to deploy on Thursday, but super excited to be leaving scap behind \o/ Will there be git-deploy gurus on call or generally available during upcoming deployment windows?
On Tue, Jan 15, 2013 at 4:51 PM, Rob Lanphier robla@wikimedia.org wrote:
Hi Arthur,
Sorry for the delayed reply. Comments below:
On Tue, Jan 15, 2013 at 12:07 PM, Arthur Richards arichards@wikimedia.org wrote:
I am scheduled to do a deployment (to all wikis) for MobileFrontend this Thursday afternoon:
- I presume scap/sync-blah will still work for 1.21wmf7 - can you please
confirm?
After tomorrow's deployment, assuming it goes well, it's going to be all git-deploy, all the time, and scap/sync-file will go away.
- Will I *have* to use git-deploy for 1.21wmf8 or will scap/sync-blah
still
work?
You'll need to use git-deploy.
And to reiterate my question from a few days ago re the deployment freeze next week:
- During the 'deployment freeze', in the event that someone needs to
deploy
an emergency fix, what do we do and who do we need to communicate with?
Use #wikimedia-operations. If no one is around, and it is an emergency, go ahead and make the fix, and note it in the logs.
Will scap/deploy-whatever still be functional, or will gti-deploy be the de-facto deployment method at that point?
git-deploy is going to be it.
Rob
wikitech-l@lists.wikimedia.org