Firefox/Channels/Postmortem/62: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
m (→‎Challenges: add a bit more detail)
(add notes from etherpad)
 
Line 1: Line 1:
Notes for 62 post mortem
Notes for 62 post mortem


10:00am PDT Tuesday (after the Channel Meeting)
10:00am PDT Tuesday September 18 (after the Channel Meeting)
Vidyo channel: release coordination
Vidyo channel: ReleaseCoordination
IRC: #release-drivers
IRC: #release-drivers


== Some stats ==
== Some stats ==


* around 340 uplifts to beta62
* uplifts during beta: https://mzl.la/2MmhuTq
* uplifts during RC week: https://mzl.la/2oYt6Ts


== What went well? ==
== What went well? ==


* WNP live and tested during RC week
* Long beta cycle but still a very stable release post-launch. Bravo! \o/
* Decision to add preference for AutoConfig sandboxing in beta (Talos as well as customers)
* When Liz went off, other release managers have been able to pick up the work without severe issues.
(she should be back next week btw)


== Challenges ==
== Challenges ==
* Long beta cycle (like we had with aurora a while back now)
** More code lands in nightly, more codes, more risks (more risk for the next release?)
** Hard on the release manager owners
** [tom] from QA perspective they liked the longer cycle
** [pascalc] Beta cycle occurred during a time when many people were on PTO (summer holidays)
* Fennec Spike in Beta 19 - {{Bug|1485441}}
* Fennec Spike in Beta 19 - {{Bug|1485441}}
** Filed when it was detected and nominated for tracking
** Filed when it was detected and nominated for tracking
** (Ritu) This is a good example of pushing RC build to 5% of release and helping catch critical problems.
** (Ritu) This is a good example of pushing RC build to 5% of release and helping catch critical problems.
** Some discussion about how we can obtain this type of device and whether the service that we are using in October would have this chipset
* ESR52 -> ESR60 migration issues
** Unable to watershed for older CPU detection - {{Bug|1415336}}
** Users of older JAWS versions not e10s-compatible were migrated - {{Bug|1489605}}
** Tom Prince: releng & release mgmt should meet to discuss about watershed before the release
*** We (RM) should add this to the checklists
** RC respin due to problems generating partial MARs - {{Bug|1481121}}, {{Bug|1488127}}
*** by the time of 68.1esr, even if we don't enable updates from 60.x until 68.2, we should have them ready and tested
** Should we call the .0 and .1 ESR releases Betas to see if enterprise users are more inclined to test them before we ship .2?
* Acer partner build bricking - {{Bug|1488934}}
** Not caught before shipping due to difficulty in testing these builds
** This was an extreme edge case scenario (unpacked XPIs being used, support for those dropped in 62), requires migrating from 58 -> 62
** Working with Softvision to add tests
** Consider whether partners are impacted when we announce future structural/build/config changes
* Search langpack issue - {{Bug|1489820}}
** Langpacks are not easy to use on beta (xpi subdir). Can we do anything to make that easier?
** beta langpacks are uploaded to AMO but unlisted so they don't get updates.  No "beta" update channel on AMO

Latest revision as of 17:59, 18 September 2018

Notes for 62 post mortem

10:00am PDT Tuesday September 18 (after the Channel Meeting) Vidyo channel: ReleaseCoordination IRC: #release-drivers

Some stats

What went well?

  • WNP live and tested during RC week
  • Long beta cycle but still a very stable release post-launch. Bravo! \o/
  • Decision to add preference for AutoConfig sandboxing in beta (Talos as well as customers)
  • When Liz went off, other release managers have been able to pick up the work without severe issues.

(she should be back next week btw)

Challenges

  • Long beta cycle (like we had with aurora a while back now)
    • More code lands in nightly, more codes, more risks (more risk for the next release?)
    • Hard on the release manager owners
    • [tom] from QA perspective they liked the longer cycle
    • [pascalc] Beta cycle occurred during a time when many people were on PTO (summer holidays)
  • Fennec Spike in Beta 19 - bug 1485441
    • Filed when it was detected and nominated for tracking
    • (Ritu) This is a good example of pushing RC build to 5% of release and helping catch critical problems.
    • Some discussion about how we can obtain this type of device and whether the service that we are using in October would have this chipset
  • ESR52 -> ESR60 migration issues
    • Unable to watershed for older CPU detection - bug 1415336
    • Users of older JAWS versions not e10s-compatible were migrated - bug 1489605
    • Tom Prince: releng & release mgmt should meet to discuss about watershed before the release
      • We (RM) should add this to the checklists
    • RC respin due to problems generating partial MARs - bug 1481121, bug 1488127
      • by the time of 68.1esr, even if we don't enable updates from 60.x until 68.2, we should have them ready and tested
    • Should we call the .0 and .1 ESR releases Betas to see if enterprise users are more inclined to test them before we ship .2?


  • Acer partner build bricking - bug 1488934
    • Not caught before shipping due to difficulty in testing these builds
    • This was an extreme edge case scenario (unpacked XPIs being used, support for those dropped in 62), requires migrating from 58 -> 62
    • Working with Softvision to add tests
    • Consider whether partners are impacted when we announce future structural/build/config changes
  • Search langpack issue - bug 1489820
    • Langpacks are not easy to use on beta (xpi subdir). Can we do anything to make that easier?
    • beta langpacks are uploaded to AMO but unlisted so they don't get updates. No "beta" update channel on AMO