Attempting to "upgrade" from `main` to 4.x should abort; Django does
not prevent running old code against the new database (though it
likely errors at runtime), and `./manage.py migrate` from the old
version during the "upgrade" does not downgrade the database, since
the migrations are entirely missing in that directory, so don't get
reversed.
Compare the list of applied migrations to the list of on-disk
migrations, and abort if there are applied migrations which are not
found on disk.
Fixes: #19284.
Services like go-camo and smokescreen are not stopped in stop-server,
since they are upgraded and restarted by puppet application. As such,
they also do not appear in start-server, despite the server relying on
them to be running to function properly.
Ensure those services are started, by starting them in start-server,
if they are configured in supervisor on the host.
For many uses, shelling out to `supervisorctl` is going to produce
better error messages. However, for instances where we wish to parse
the output of `supervisorctl`, using the API directly is less brittle.
The RabbitMQ docs state ([1]):
RabbitMQ nodes and CLI tools (e.g. rabbitmqctl) use a cookie to
determine whether they are allowed to communicate with each
other. [...] The cookie is just a string of alphanumeric
characters up to 255 characters in size. It is usually stored in a
local file.
...and goes on to state (emphasis ours):
If the file does not exist, Erlang VM will try to create one with
a randomly generated value when the RabbitMQ server starts
up. Using such generated cookie files are **appropriate in
development environments only.**
The auto-generated cookie does not use cryptographic sources of
randomness, and generates 20 characters of `[A-Z]`. Because of a
semi-predictable seed, the entropy of this password is thus less than
the idealized 26^20 = 94 bits of entropy; in actuality, it is 36 bits
of entropy, or potentially as low as 20 if the performance of the
server is known.
These sizes are well within the scope of remote brute-force attacks.
On provision, install, and upgrade, replace the default insecure
20-character Erlang cookie with a cryptographically secure
255-character string (the max length allowed).
[1] https://www.rabbitmq.com/clustering.html#erlang-cookie
5c450afd2d, in ancient history, switched from `check_call` to
`check_output` and throwing away its result.
Use check_call, so that we show the steps to (re)starting the server.
This is required in order to lock down the RabbitMQ port to only
listen on localhost. If the nodename is `rabbit@hostname`, in most
circumstances the hostname will resolve to an external IP, which the
rabbitmq port will not be bound to.
Installs which used `rabbit@hostname`, due to RabbitMQ having been
installed before Zulip, would not have functioned if the host or
RabbitMQ service was restarted, as the localhost restrictions in the
RabbitMQ configuration would have made rabbitmqctl (and Zulip cron
jobs that call it) unable to find the rabbitmq server.
The previous commit ensures that configure-rabbitmq is re-run after
the nodename has changed. However, rabbitmq needs to be stopped
before `rabbitmq-env.conf` is changed; we use an `onlyif` on an `exec`
to print the warning about the node change, and let the subsequent
config change and notify of the service and configure-rabbitmq to
complete the re-configuration.
`/etc/rabbitmq/rabbitmq-env.conf` sets the nodename; anytime the
nodename changes, the backing database changes, and this requires
re-creating the rabbitmq users and permissions.
Trigger this in puppet by running configure-rabbitmq after the file
changes.
This reverts commit 889547ff5e. It is
unused in the Docker container, as the configurtaion of the `zulip`
user in the rabbitmq node is done via environment variables. The
Zulip host in that context does not have `rabbitmqctl` installed, and
would have needed to know the Erlang cookie to be able to run these
commands.
This addresses the problems mentioned in the previous commit, but for
existing installations which have `authenticator = standalone` in
their configurations.
This reconfigures all hostnames in certbot to use the webroot
authenticator, and attempts to force-renew their certificates.
Force-renewal is necessary because certbot contains no way to merely
update the configuration. Let's Encrypt allows for multiple extra
renewals per week, so this is a reasonable cost.
Because the certbot configuration is `configobj`, and not
`configparser`, we have no way to easily parse to determine if webroot
is in use; additionally, `certbot certificates` does not provide this
information. We use `grep`, on the assumption that this will catch
nearly all cases.
It is possible that this will find `authenticator = standalone`
certificates which are managed by Certbot, but not Zulip certificates.
These certificates would also fail to renew while Zulip is running, so
switching them to use the Zulip webroot would still be an improvement.
Fixes#20593.
Installing certbot with --method=standalone means that the
configuration file will be written to assume that the standalone
method will be used going forward. Since nginx will be running,
attempts to renew the certificate will fail.
Install a temporary self-signed certificate, just to allow nginx to
start, and then follow up (after applying puppet to start nginx) with
the call to setup-certbot, which will use the webroot authenticator.
The `setup-certbot --method=standalone` option is left intact, for use
in development environments.
Fixes part of #20593; it does not address installs which were
previously improperly configured with `authenticator = standalone`.
As a consequence:
• Bump minimum supported Python version to 3.7.
• Move Vagrant environment to Debian 10, which has Python 3.7.
• Move CI frontend tests to Debian 10.
• Move production build test to Debian 10.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
This catches nine functional indexes that the previous query didn’t:
upper_preregistration_email_idx
upper_stream_name_idx
upper_subject_idx
upper_userprofile_email_idx
zerver_message_recipient_upper_subject
zerver_mutedtopic_stream_topic
zerver_stream_realm_id_name_uniq
zerver_userprofile_realm_id_delivery_email_uniq
zerver_userprofile_realm_id_email_uniq
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Restarting the uwsgi processes by way of supervisor opens a window
during which nginx 502's all responses. uwsgi has a configuration
called "chain reloading" which allows for rolling restart of the uwsgi
processes, such that only one process at once in unavailable; see
uwsgi documentation ([1]).
The tradeoff is that this requires that the uwsgi processes load the
libraries after forking, rather than before ("lazy apps"); in theory
this can lead to larger memory footprints, since they are not shared.
In practice, as Django defers much of the loading, this is not as much
of an issue. In a very basic test of memory consumption (measured by
total memory - free - caches - buffers; 6 uwsgi workers), both
immediately after restarting Django, and after requesting `/` 60 times
with 6 concurrent requests:
| Non-lazy | Lazy app | Difference
------------------+------------+------------+-------------
Fresh | 2,827,216 | 2,870,480 | +43,264
After 60 requests | 3,332,284 | 3,409,608 | +77,324
..................|............|............|.............
Difference | +505,068 | +539,128 | +34,060
That is, "lazy app" loading increased the footprint pre-requests by
43MB, and after 60 requests grew the memory footprint by 539MB, as
opposed to non-lazy loading, which grew it by 505MB. Using wsgi "lazy
app" loading does increase the memory footprint, but not by a large
percentage.
The other effect is that processes may be served by either old or new
code during the restart window. This may cause transient failures
when new frontend code talks to old backend code.
Enable chain-reloading during graceful, puppetless restarts, but only
if enabled via a zulip.conf configuration flag.
Fixes#2559.
[1]: https://uwsgi-docs.readthedocs.io/en/latest/articles/TheArtOfGracefulReloading.html#chain-reloading-lazy-apps
On a system where ‘apt-get update’ has never been run, ‘apt-cache
policy’ may show no repositories at all. Try to correct this with
‘apt-get update’ before giving up.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
If nginx was already installed, and we're using the webroot method of
initializing certbot, nginx needs to be reloaded. Hooks in
`/etc/letsencrypt/renewal-hooks/deploy/` do not run during initial
`certbot certonly`, so an explicit reload is required.
The certbot package installs its own systemd timer (and cron job,
which disabled itself if systemd is enabled) which updates
certificates. This process races with the cron job which Zulip
installs -- the only difference being that Zulip respects the
`certbot.auto_renew` setting, and that it passes the deploy hook.
This means that occasionally nginx would not be reloaded, when the
systemd timer caught the expiration first.
Remove the custom cron job and `certbot-maybe-renew` script, and
reconfigure certbot to always reload nginx after deploying, using
certbot directory hooks.
Since `certbot.auto_renew` can't have an effect, remove the setting.
In turn, this removes the need for `--no-zulip-conf` to
`setup-certbot`. `--deploy-hook` is similarly removed, as running
deploy hooks to restart nginx is now the default; pass
`--no-directory-hooks` in standalone mode to not attempt to reload
nginx. The other property of `--deploy-hook`, of skipping symlinking
into place, is given its own flog.
We've had a number of unhappy reports of upgrades failing due to
webpack requiring too much memory. While the previous commit will
likely fix this issue for everyone, it's worth improving the error
message for failures here.
We avoid doing the stop+retry ourselves, because that could cause an
outage in a production system if webpack fails for another reason.
Fixes#20105.
Since the upgrade to Webpack 5, we've been seeing occasional reports
that servers with roughly 4GiB of RAM were getting OOM kills while
running webpack.
Since we can't readily optimize the memory requirements for webpack
itself, we should raise the RAM requirements for doing the
lower-downtime upgrade strategy.
Fixes#20231.
Stopping both `zulip-tornado` and `zulip-tornado:*` causes errors on
deploys with tornado sharding, as the plain `zulip-tornado` service
does not exist.
Pass `zulip-tornado:*`, which matches both plain `zulip-tornado`, as
well as the sharded `zulip-tornado:zulip-tornado-port-9800` cases.
The `analyze_new_cluster.sh` script output by `pg_upgrade` just runs
`vacuumdb --all --analyze-in-stages`, which runs three passes over the
database, getting better stats each time. Each of these passes is
independent; the third pass does not require the first two.
`--analyze-in-stages` is only provided to get "something" into the
database, on the theory that it could then be started and used. Since
we wait for all three passes to complete before starting the database,
the first two passes add no value.
Additionally, PosttgreSQL 14 and up stop writing the
`analyze_new_cluster.sh` script as part of `pg_upgrade`, suggesting
the equivalent `vacuumdb --all --analyze-in-stages` call instead.
Switch to explicitly call `vacuumdb --all --analyze-only`, since we do
not gain any benefit from `--analyze-in-stages`. We also enable
parallelism, with `--jobs 10`, in order to analyze up to 10 tables in
parallel. This may increase load, but will accelerate the upgrade
process.
scripts.lib.node_cache expects Yarn to be in /srv/zulip-yarn, so if
it’s installed somewhere else, even if it’s the right version, we need
to reinstall it.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
It recently started failing on Debian 10 (buster). We immediately
follow this by replacing these packages with our own versions from
pip.txt, anyway.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
The support for bullseye was added in #17951
but it was not documented as bullseye was
frozen and did not have proper configuration
files, hence wasn't documented.
Since now bullseye is released as a stable
version, it's support can be documented.
The usual output from this command looks like
Notice: Compiled catalog for localhost in environment production in 2.33 seconds
Notice: /Stage[main]/Zulip::Apt_repository/Exec[setup_apt_repo]/returns: current_value 'notrun', should be ['0'] (noop)
Notice: Class[Zulip::Apt_repository]: Would have triggered 'refresh' from 1 event
Notice: Stage[main]: Would have triggered 'refresh' from 1 event
Notice: Applied catalog in 1.20 seconds
which doesn’t seem abnormally alarming, and hiding it makes failures
harder to diagnose.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
We previously used `zulip-puppet-apply` with a custom config file,
with an updated PostgreSQL version but more limited set of
`puppet_classes`, to pre-create the basic settings for the new cluster
before running `pg_upgradecluster`.
Unfortunately, the supervisor config uses `purge => true` to remove
all SUPERVISOR configuration files that are not included in the puppet
configuration; this leads to it removing all other supervisor
processes during the upgrade, only to add them back and start them
during the second `zulip-puppet-apply`.
It also leads to `process-fts-updates` not being started after the
upgrade completes; this is the one supervisor config file which was
not removed and re-added, and thus the one that is not re-started due
to having been re-added. This was not detected in CI because CI added
a `start-server` command which was not in the upgrade documentation.
Set a custom facter fact that prevents the `purge` behaviour of the
supervisor configuration. We want to preserve that behaviour in
general, and using `zulip-puppet-apply` continues to be the best way
to pre-set-up the PostgreSQL configuration -- but we wish to avoid
that behaviour when we know we are applying a subset of the puppet
classes.
Since supervisor configs are no longer removed and re-added, this
requires an explicit start-server step in the instructions after the
upgrades complete. This brings the documentation into alignment with
what CI is testing.
These changes are all independent of each other; I just didn’t feel
like making dozens of commits for them.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
For our marketing emails, we want a width that's more appropriate for
newsletter context, vs. the narrow emails we use for transactional
content.
I haven't figured out a cleaner way to do this than duplicating most
of email_base_default.source.html. But it's not a big deal to
duplicate, since we've been changing that base template only about
once a year.
The script is added to upgrade steps for 20.04 and Buster because
those are the upgrades that cross glibc 2.28, which is most
problematic. It will also be called out in the upgrade notes, to
catch those that have already done that upgrade.
Running `supervisorctl stop` or `supervisorctl restart` on a process
name which is not known is an error:
```
$ supervisorctl stop nonexistent-process
nonexistent-process: ERROR (no such process)
$ echo $?
1
```
ef6d0ec5ca moved
zulip_deliver_scheduled_* out of the `workers:` group. Since upgrades
run `stop-server` before applying puppet, the list of processes at
that time is from the previous version of Zulip, so may not have the
new `zulip_deliver_scheduled_*` names -- and the `stop-server` will
hence fail.
If the upgrade is not applying puppet, it will `restart-server`. At
that point, the old names will still be in the configuration, so
relying on the current `superisorctl status` is the best gauge of what
exists to restart.
In short, only ever stop/start/restart the `zulip_deliver_scheduled_*`
processes if `supervisorctl status` knows about them already.
Nonexistent processes and groups passed to `supervisortctl status` are
printed to STDOUT as follows:
```
$ supervisorctl status zulip-django nonexistent-process nonexistent-group:*
nonexistent-process: ERROR (no such process)
nonexistent-group: ERROR (no such group)
zulip-django RUNNING pid 16043, uptime 17:31:31
```
On supervisor 4 and above, this exits with an exit code of 4;
previously, it returned exit code 0. Ubuntu 18.04 has version 3.3.1,
and Ubuntu 20.04 has version 4.1.0.
Skip any lines with `ERROR (no such ...)`, and accept exit code 4 from
`supervisorctl status`.
Mypy can’t follow absolute imports based on directories other than the
root. This was hiding some type errors due to ignore_missing_imports.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
With two space-separated classes in `puppet_classes`, the second one
is silently ignored. With three of more, puppet generates the
following very opaque error message:
```
Error: Could not parse for environment production: This
Name has no effect. A value was produced and then forgotten (one or
more preceding expressions may have the wrong form)
```
Catch when this has happened, and give an error message to the user.
Fixes#18992.
This command is part of a statsd infrastructure that we stopped
supporting years ago. Its only purpose for some time has been to
provide sample code for how the restart script might trigger a
notification to a graphing system, which doesn't justify maintaining
it.
Fixes part of #18898.
The current `upgrade-zulip` and `upgrade-zulip-from-git`
bash scripts exit with a zero status even if the
upgrade commands exit with a non-zero status.
Hence add `set -e` command which exits the script with
the same status as the non-zero command.
For pipe commands however, the net status of a command
is the status of the last command, hence if the other parts
fail, the net status is only determined by the last command.
This is the case with our main /lib/upgrade-zulip* command
in the scripts whose status is determined by the `tee` command
instead. Hence add a small condition to get the status of the
actual upgrade command and exit the script if it fails with
a non-zero command.
We also check whether the script is being run as root, matching the
install script logic.
This parameter is somewhat useful, and adding this also fixes a
regression where purge-old-deployments would crash since the changes
around c5580607a7 because of
inconsistent supported args lists.
Fixes#16659.
If the server is behind a reverse proxy with http_only=True, the
requests made by email-mirror-postfix need to use http, as https
doesn't work.
Staging and other hosts that are `zulip::app_frontend_base` but not
`zulip::app_frontend_once` do not have a
/etc/supervisor/conf.d/zulip/zulip-once.conf and as such do not have
`zulip_deliver_scheduled_emails` or `zulip_deliver_scheduled_messages`
and thus supervisor will fail to reload.
Making the contents of `zulip-workers` contingent on if the server is
_also_ a `-once` server is complicated, and would involve using Concat
fragments, which severely limit readability.
Instead, expel those two from `zulip-workers`; this is somewhat
reasonable, since they are use an entirely different codepath from
zulip_events_*, using the database rather than RabbitMQ for their
queuing.
We currently run the `clean_unused_caches.py` as a
script to clean the unused caches.
This commit replaces that with
`clean_unused_caches.main` function as it would be
faster.
This commit will allow us to pass the arguments in the
'clean...' functions when calling the `main` function (in
`provision`). It also changes args parsing
function location to `if __name__ == "__main__"` block as
we wouldn't need it to parse args when we call the
function.
We convert the `clean-unused-caches` script to a
python file so we can run it in provision by importing it
instead of running the script, hence saving some time.
Anders Kaseorg
Tim Abbott
Alex Vandiver
rht
Gaurav Pandey
manavdesai27
Mateusz Mandera
Riken Shah