Web update cannot continue after long backup

[blizzz]

Sometimes I am stuck at this position when using web updater:

And Retry update would fail, because the reason is that the updater secret is meanwhile outdated:

That is the same secret with which the previous "step 4" (download) succeeded. The previous backup step took 15.3s.

Something must have reset or deleted the updater secret in this process.

Opening the admin overview page and from there opening the updater again, let's me finish the update.

[joshtrichards]

Something must have reset or deleted the updater secret in this process.

Did you happen to notice whether the updater.secret truly changed in your config.php in between those steps (but before coming back in via the Overview screen to reset)?

I looked into this recently based on a handful of mysterious reports from the forums (where I was able to gather enough log evidence to convince myself something like this was going on... at least in more than just a single isolated environment).

That led me to collect this: nextcloud/server#45714

Given the Updater itself is just a consumer of the configured updater.secret my assumption has been either:

• it's changing when it's not supposed to be (in Server via the updatenotification app)
• we're losing track of the current/correct one in the Updater itself or the browser side somehow (your experience suggests it's probably not the browser side)

[blizzz]

Sorry, did not come round back to it, yet.

Did you happen to notice whether the updater.secret truly changed in your config.php in between those steps (but before coming back in via the Overview screen to reset)?

I checked before-after only.

That led me to collect this: nextcloud/server#45714

👍

we're losing track of the current/correct one in the Updater itself or the browser side somehow (your experience suggests it's probably not the browser side)

ack

it's changing when it's not supposed to be (in Server via the updatenotification app)

That's what I am thinking though I could not find a very obvious place so far.

[blizzz]

for the record, i could not repro lately (while debug patch was applied).

[michaellass]

I faced this issue a couple of times now, last time when updating from 29.0.5 to 29.0.6. The update from 29.0.6 to 29.0.7 now went smoothly without this issue showing up. Was there a change that may have fixed this issue?

[joshtrichards]

No, not specifically.

[michaellass]

No, not specifically.

I can now confirm that the update to 29.0.7 was a fluke. The issue appeared again when updating to 29.0.8.

[joshtrichards]

Found the cause:

https://github.com/nextcloud/server/blob/e17f011f08cb4529ae19ff8637a49b7f4e89f216/apps/updatenotification/lib/BackgroundJob/ResetToken.php#L43-L45

We're not checking the difference since nextcloud/server#43967 was merged. It has unit test coverage, but they weren't sufficient to catch this because the hypothetical dates happened to be small enough to fit under the radar.

Fix pending. See nextcloud/server#49578