Page MenuHomeSoftware Heritage

production: Deploy swh.loader.mercurial v2.1.0
Closed, ResolvedPublic

Description

  • Upgrade swh.storage to the latest
  • Restart storage services
  • Upgrade workers and restart the services swh-worker@loader_mercurial and swh-worker@loader_high_priority.

At the end of it all though, the final production check end-to-end for mercurial origin should go green.

Event Timeline

ardumont triaged this task as Unbreak Now! priority.
ardumont created this task.
ardumont updated the task description. (Show Details)

At the end of it all though, the final production check end-to-end for mercurial origin should go green.

It's working but the check does not pass green [1]. As far as i could tell, the unsuccessful
event [1] is seen as failure by the check.

Unsuccessful now because one load was successful first (through swh loader run on one
worker because no logs were seen and i grew bored looking for it).

[1] As can be seen by the extid_version set to 1. This means the hg loader did its job
(around ~1100 extids for the tested origin). There are more here because the sourceforge
mercurial origins also started.

17:39:06 softwareheritage@belvedere:5432=> select now(), extid_type, extid_version, count(*) from extid group by extid_type, extid_version order by extid_version desc;
+-------------------------------+-------------------------+---------------+----------+
|              now              |       extid_type        | extid_version |  count   |
+-------------------------------+-------------------------+---------------+----------+
| 2021-07-29 15:38:42.289853+00 | hg-nodeid               |             1 |    32774 |
| 2021-07-29 15:38:42.289853+00 | cran-sha256             |             0 |    30131 |
| 2021-07-29 15:38:42.289853+00 | dsc-sha256              |             0 |    67753 |
| 2021-07-29 15:38:42.289853+00 | hg-nodeid               |             0 | 19587036 |
| 2021-07-29 15:38:42.289853+00 | npm-archive-sha1        |             0 | 19461834 |
| 2021-07-29 15:38:42.289853+00 | package-manifest-sha256 |             0 |     8391 |
| 2021-07-29 15:38:42.289853+00 | pypi-archive-sha256     |             0 |  2519490 |
| 2021-07-29 15:38:42.289853+00 | subresource-integrity   |             0 |    68712 |
+-------------------------------+-------------------------+---------------+----------+
(8 rows)

Time: 19726.983 ms (00:19.727)
ardumont changed the task status from Open to Work in Progress.Jul 29 2021, 5:40 PM
ardumont moved this task from Weekly backlog to in-progress on the System administration board.

It's working but the check does not pass green [1]. As far as i could tell, the unsuccessful
event [1] is seen as failure by the check.

That was not what i hypothesized.
There is a glitch somewhere which looks like a race condition in the status updates
between the scheduler db and the swh-web db.

But let's keep that for another time.

It's finally back to green as initially expected [1]

18:16 <+swhbot> icinga RECOVERY: service production Check save-code-now hgview with type hg end-to-end on pergamon.softwareheritage.org is OK: SAVECODENOW OK - Save code now request for origin ('hg', 'https://foss.heptapod.net/mercurial/hgview') took 75.73s and succeeded.

Closing now.