WebDevITQAMtg09102010: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
No edit summary
No edit summary
 
Line 1: Line 1:
* Ways to get Webdev/IT/WebQA on the same page, concerning versions of libraries and binaries (PHP/Python/MySQL, etc.)
*Ways to get Webdev/IT/WebQA on the same page, concerning versions of libraries and binaries (PHP/Python/MySQL, etc.)  
** Desired/necessary level of formality: pre-release checklist? What determines "sign off"?
**Desired/necessary level of formality: pre-release checklist? What determines "sign off"?  
*** Day before, pre-release meetings
***Day before, pre-release meetings  
*** Webdev lead makes the call and does go/no-go decision
***Webdev lead makes the call and does go/no-go decision  
* Where the responsibility of application/site performance and stress-testing lies, and when it's needed
*Where the responsibility of application/site performance and stress-testing lies, and when it's needed  
** Still lies with IT and WebDev
**Still lies with IT and WebDev  
* How we can get better at scheduling releases (https://bugzilla.mozilla.org/show_bug.cgi?id=587824 comes to mind)? What's the process? Do we need rules about how many combined large (AMO/SUMO, etc.) and small (Personas, Input) releases we do in a week?
*How we can get better at scheduling releases (https://bugzilla.mozilla.org/show_bug.cgi?id=587824 comes to mind)? What's the process? Do we need rules about how many combined large (AMO/SUMO, etc.) and small (Personas, Input) releases we do in a week?  
** Leave it up to WebDev/WebQA to gauge
**Leave it up to WebDev/WebQA to gauge  
** OK with 4pm releases, can't do earlier than 4pm
**OK with 4pm releases, can't do earlier than 4pm  
* We launched our last AMO release with a brand-new Zeus cluster, but didn't test it prior to launch; maybe Wil knew, but could QA have tested, earlier? (There were no problems, but there could've been.)
*We launched our last AMO release with a brand-new Zeus cluster, but didn't test it prior to launch; maybe Wil knew, but could QA have tested, earlier? (There were no problems, but there could've been.)  
* How do we ensure we have user-data backed up, pre-push?
*How do we ensure we have user-data backed up, pre-push?  
* If something goes wrong during a push, what are the criteria (I realize this is hard to talk to, in terms of specifics) by which we decide to roll back? What's the allowed user-impacting downtime? Is it the outage window? If we think we can solve a problem, is it fine to extend past the outage window? Do we need to message our users further, with an ETA, etc.?
*If something goes wrong during a push, what are the criteria (I realize this is hard to talk to, in terms of specifics) by which we decide to roll back? What's the allowed user-impacting downtime? Is it the outage window? If we think we can solve a problem, is it fine to extend past the outage window? Do we need to message our users further, with an ETA, etc.?  
** we don't truly "panic" until the end of the outage window
**we don't truly "panic" until the end of the outage window


* Outcomes/followups:
*Outcomes/followups:  
** mrz to try to ensure that push bugs have owners earlier (mrz)
**try to ensure that push bugs have owners earlier ('''mrz''')  
** chris is rolling out a security-inventory system (mrz)
**chris is rolling out a security-inventory system ('''mrz''')  
*** put the app behind VPN and have WebDev learn how to use it, and do inventories (mrz/morgamic)
***put the app behind VPN and have WebDev learn how to use it, and do inventories ('''mrz/morgamic''')  
** case-by-case basis for sites that need to graduate from VM -> prod (webqa/webdev)
**case-by-case basis for sites that need to graduate from VM -> prod ('''webqa/webdev''')  
** set up cron jobs to copy prod DBs -> staging (webqa/webdev to raise issue to IT when it arises)
**set up cron jobs to copy prod DBs -> staging ('''webqa/webdev''' to raise issue to IT when it arises)  
** lead developer should know their schemas/run schematic -- to make sure that the schemas are the same between prod/staging (can diff the schemas) (webdev)
**lead developer should know their schemas/run schematic -- to make sure that the schemas are the same between prod/staging (can diff the schemas) ('''webdev''')  
** come up with more useful or app-specific outage pages, and a way/person to update that content with a helpful ETA, etc. (owner needed)
**come up with more useful or app-specific outage pages, and a way/person to update that content with a helpful ETA, etc. ('''''owner needed''''')  
** try switching the docroot and the DB on a virtual IP for a small site (mrz/webdev)
**try switching the docroot and the DB on a virtual IP for a small site ('''mrz/webdev''')  
** can do pushes Tuesday, Wednesday, Thursday (Wed. only if needed)
**can do pushes Tuesday, Wednesday, Thursday (Wed. only if needed)  
** small content changes can go whenever
**small content changes can go whenever  
** webdev to enable read-only mode on more sites (webdev)
**enable read-only mode on more sites ('''webdev''')

Latest revision as of 00:18, 11 September 2010

  • Ways to get Webdev/IT/WebQA on the same page, concerning versions of libraries and binaries (PHP/Python/MySQL, etc.)
    • Desired/necessary level of formality: pre-release checklist? What determines "sign off"?
      • Day before, pre-release meetings
      • Webdev lead makes the call and does go/no-go decision
  • Where the responsibility of application/site performance and stress-testing lies, and when it's needed
    • Still lies with IT and WebDev
  • How we can get better at scheduling releases (https://bugzilla.mozilla.org/show_bug.cgi?id=587824 comes to mind)? What's the process? Do we need rules about how many combined large (AMO/SUMO, etc.) and small (Personas, Input) releases we do in a week?
    • Leave it up to WebDev/WebQA to gauge
    • OK with 4pm releases, can't do earlier than 4pm
  • We launched our last AMO release with a brand-new Zeus cluster, but didn't test it prior to launch; maybe Wil knew, but could QA have tested, earlier? (There were no problems, but there could've been.)
  • How do we ensure we have user-data backed up, pre-push?
  • If something goes wrong during a push, what are the criteria (I realize this is hard to talk to, in terms of specifics) by which we decide to roll back? What's the allowed user-impacting downtime? Is it the outage window? If we think we can solve a problem, is it fine to extend past the outage window? Do we need to message our users further, with an ETA, etc.?
    • we don't truly "panic" until the end of the outage window
  • Outcomes/followups:
    • try to ensure that push bugs have owners earlier (mrz)
    • chris is rolling out a security-inventory system (mrz)
      • put the app behind VPN and have WebDev learn how to use it, and do inventories (mrz/morgamic)
    • case-by-case basis for sites that need to graduate from VM -> prod (webqa/webdev)
    • set up cron jobs to copy prod DBs -> staging (webqa/webdev to raise issue to IT when it arises)
    • lead developer should know their schemas/run schematic -- to make sure that the schemas are the same between prod/staging (can diff the schemas) (webdev)
    • come up with more useful or app-specific outage pages, and a way/person to update that content with a helpful ETA, etc. (owner needed)
    • try switching the docroot and the DB on a virtual IP for a small site (mrz/webdev)
    • can do pushes Tuesday, Wednesday, Thursday (Wed. only if needed)
    • small content changes can go whenever
    • enable read-only mode on more sites (webdev)