Zulip 13.0: What's New and How to Upgrade Safely
Zulip Server 13.0 came out on November 12, 2026, and two quick point releases followed within a day, so 13.2 is the version to install. For most teams it's a pleasant upgrade: better ways to find related conversations, a date search, and a cleaner interface.
For bigger instances there's one thing to plan for before you run it. The upgrade rebuilds four indexes on Zulip's largest table, and that needs both time and spare disk space. Here's what's new, how to check whether that affects you, and how to upgrade a standard install or a Docker setup.
What's New in 13.0
The highlights from the Zulip changelog:
- Topic connections. A new View links option in the topic menu shows which other conversations link to a topic, and which ones it links to. Useful when a decision is spread over three threads.
- A
date:search operator that jumps to messages around a given day. - Topic search. A new Search topics button next to the channel you're viewing, plus topic search across all channels from the left sidebar.
- Hide link previews one message at a time, from the message menu or with
Shift+L. - A "next unread" button at the end of a conversation.
- Mobile notifications on by default for new subscribers to channels where an admin enables it.
- Visual updates, including redesigned modal buttons and underlined links in messages.
The Part to Plan For: Four Index Rebuilds
The upgrade notes say 13.0 rebuilds four indexes on the zerver_message table, which holds every message your organization has ever sent. On a small instance that's quick. On a large one, Zulip warns it "will take a while", and you need enough free disk space for a second copy of those indexes, because the old ones are only dropped once the new ones are built.
There's a second change that matters here. Migrations that add or remove indexes no longer use PostgreSQL's CONCURRENTLY option by default, so they lock the table while they run. The standard upgrade stops the server before migrating anyway, so it doesn't change anything there. It only matters if you run migrations while Zulip is up, and in that case you set this in /etc/zulip/settings.py first:
MIGRATIONS_ADD_REMOVE_INDEXES_CONCURRENTLY = True
How Much Space Do You Need?
Check the total size of the message table's indexes. That's more than you strictly need, since only four of them get rebuilt, which makes it a safe upper bound. On a standard install:
sudo -u zulip psql -c "SELECT pg_size_pretty(pg_indexes_size('zulip.zerver_message'));"
On docker-zulip, run the same query in the database container:
docker compose exec database psql -U zulip zulip \
-c "SELECT pg_size_pretty(pg_indexes_size('zulip.zerver_message'));"
If the table isn't found, drop the zulip. prefix. Then compare the result with df -h on the disk that holds PostgreSQL's data. If free space is close to that number, add space before upgrading.
Two Smaller Breaking Changes
- Global time syntax.
<time:…>now only accepts ISO 8601 timestamps. UNIX timestamps and human-readable dates are no longer converted, so update any bots or templates that send them. - Heroku integration. It now only accepts Heroku app webhooks, and legacy deploy hooks are gone. If you used one, set the integration up again.
Upgrading a Standard Install
Back up first. Zulip's own backup tool covers the database, uploads and settings:
su zulip -c '/home/zulip/deployments/current/manage.py backup --output=/home/zulip/before-13.tar.gz'
Then, as root, download and run the upgrade:
curl -fLO https://download.zulip.com/server/zulip-server-latest.tar.gz
/home/zulip/deployments/current/scripts/upgrade-zulip zulip-server-latest.tar.gz
The upgrade guide says downtime is normally under 30 seconds unless a release includes an expensive migration. This one does, so schedule it outside working hours on a big instance. If you're jumping several versions, read the upgrade notes for every release in between.
Upgrading docker-zulip
The docker-zulip guide recommends a backup first:
docker compose exec zulip /sbin/entrypoint.sh app:backup
Then check out the release tag, which matches the image tag, and restart:
git fetch --tags
git checkout -B release 13.2-0
docker compose pull
docker compose up -d
docker compose logs -f zulip
The container runs the migrations on its first start, index rebuilds included. Leave it alone until the logs show it's finished. Your local settings in compose.override.yaml aren't touched by the checkout.
Running It on Elestio
Zulip on Elestio starts at $16/month for the VM, with automated backups included. Run the disk check above first. If you're short on space, resize the service before upgrading rather than finding out halfway through.
For a large instance, rehearse on a copy with the Clone button in your dashboard and time the upgrade there. Or let Elestio AI DevOps run it: it takes a backup before touching anything, and you watch every step.
Troubleshooting
The upgrade is taking much longer than usual. It's rebuilding the message indexes. Watch the logs rather than restarting, because interrupting a migration only makes things worse.
The upgrade fails with a disk space error. The new indexes didn't fit next to the old ones. Free up or add space, then run the upgrade again.
Users report Zulip is frozen during a live migration. You ran migrations with the server up, without setting MIGRATIONS_ADD_REMOVE_INDEXES_CONCURRENTLY = True. Index changes now lock the table by default.
A bot's timestamps show up as plain text. It's sending UNIX or human-readable times in <time:…>. Switch to ISO 8601.
Heroku notifications stopped. You were using a legacy deploy hook. Reconfigure the Heroku integration with app webhooks.
Thanks for reading ❤️ See you in the next one 👋