Docs - Stop Puma when downgrading

See merge request gitlab-org/omnibus-gitlab!3868

Add note for old restore command

See merge request gitlab-org/omnibus-gitlab!3869

Update gitlab-org/gitlab-exporter from 5.1.0 to 5.2.2

See merge request gitlab-org/omnibus-gitlab!3848

[┐∵]┘GitLab Dependency Bot authored 5 years ago and Balasankar C committed 5 years ago

And add a changelog

Bump registry to v2.7.6-gitlab from v2.7.4-gitlab

See merge request gitlab-org/omnibus-gitlab!3862

Update Mattermost to 5.18.1

See merge request gitlab-org/omnibus-gitlab!3847

Update rubygems/rubygems from 2.7.9 to 2.7.10

See merge request gitlab-org/omnibus-gitlab!3805

Fix request duration config

See merge request gitlab-org/omnibus-gitlab!3859

The application reads `Settings.gitlab.max_request_duration_seconds`

Raise unicorn memory limits

Closes #4279

See merge request gitlab-org/omnibus-gitlab!3853

Docs: auto migrate should be false in all nodes during HA upgrade

See merge request gitlab-org/omnibus-gitlab!3857

Add max_request_duration config to gitlab_rails

See merge request gitlab-org/omnibus-gitlab!3830

The base memory usage of a Unicorn worker is about 500-600 MB now. The
previous memory limits (min: 400, max: 600) were too low because Unicorn
Worker Killer picks a random number between 400 and 600 MB and will kill
the process if it happens to fall in that number. As a result, there was
a high probability that a Unicorn worker would be killed after **every**
request.

To fix this, we raise the limits to match GitLab.com: 1 GB and 1.28 GB.

We also tweak the formula for calculating number of Unicorn workers. This
increase of baseline RAM lowers the number of workers.

Closes https://gitlab.com/gitlab-org/omnibus-gitlab/issues/4279

Update https://git.code.sf.net/p/libpng/code from 1.6.35 to 1.6.37

See merge request gitlab-org/omnibus-gitlab!3800

This should simplify the upgrade process.

This option lets the rails app know how much time it has to process a
request. And make decisions accordingly.

By default the time is 90% of the timeout configured for the
worker. This gives us some time during the request to clean up if we
need to.

Setting a duration higher than the worker timeout will raise an error.

This allows for being more concise in specs that test a state when
changing multiple variables.

Link to the background migrations check documentation

Closes #4908

See merge request gitlab-org/omnibus-gitlab!3856

DJ Mountney authored 5 years ago and Achilleas Pipinellis committed 5 years ago

Ensure we mention checking background migrations when upgrading major
release version.

We are planning to bake this check into Omnibus, but we need the
documentation to exist for users still upgrading to the 12.x releases.

[ci skip]

[ci skip]

Add grpc_latency_bucket config to praefect

See merge request gitlab-org/omnibus-gitlab!3854

Add docs about upgrading Redis HA setup

See merge request gitlab-org/omnibus-gitlab!3815

Update Mattermost to 5.17.3 (GitLab 12.6)

See merge request gitlab-org/omnibus-gitlab!3846

[ci skip]

Check for puma when setting Geo primary

See merge request gitlab-org/omnibus-gitlab!3855

Update grafana-dashboards to v1.3.0 w/ Service Platform Metrics

Closes #4886

See merge request gitlab-org/omnibus-gitlab!3843

Update to latest Redis release.

See merge request gitlab-org/omnibus-gitlab!3392

[ci skip]