Built to stay up, and to recover
Portus runs on Amazon Web Services. It is checked every 30 seconds from around the world, backed up every six hours, and a release that fails its checks puts the one before it back on its own.
What Portus runs on
Every layer under Portus, from the data centre up, as it runs in production today.
Amazon Web Services
Data centres certified to ISO 27001 and SOC 1, 2 and 3, and an AWS account that holds Portus and nothing else.
Encrypted throughout
Encrypted disks, backups and snapshots, settings and keys under keys of their own, and TLS with HSTS on every connection.
Closed to the internet
The server takes web traffic and nothing else. There is no SSH: it is run through AWS Systems Manager, and every command is logged.
Services that restart themselves
Every part of Portus has its own health check and its own memory limit. A part that stops is started again on its own.
Organisations kept apart
Every row carries its organisation, and the database’s own keys stop one organisation’s records pointing at another’s.
Errors reach us at once
Every error in the app is reported to our team the moment it happens, with no one’s data in the report.
Checked every 30 seconds, from around the world
Health checkers in several locations around the world call Portus every 30 seconds. Three failures in a row mark it down, and the alarm reaches the team at once.
What raises an alarm
- The app stops answering its health checkOver2 minutes
- The server fails AWS’s status checksOver2 minutes
- The server’s CPU stays above 80%Over15 minutes
- The disk is more than 80% fullOver5 minutes
- No backup has reached the bucketOver24 hours
Each alarm reaches us by email the moment it goes off, and again when it clears.
Backed up every six hours, kept 14 days
A full backup of the database and every stored file, encrypted and kept apart from the server in Amazon S3, which stores each copy across several data centres. The server can add backups but can’t read, overwrite or delete them.
Four a day
At 00:00, 06:00, 12:00 and 18:00 UTC, which is 03:00, 09:00, 15:00 and 21:00 in Riyadh.
Kept 14 days
Each backup expires on its own after 14 days. It is encrypted with AWS KMS and accepted only over TLS.
A daily disk snapshot
At 02:00 UTC, with the last seven kept. A snapshot holds the whole disk, so the server itself can be rebuilt from it.
One command to restore
A restore is one tested command. It fetches the backup, backs up what it replaces, and brings Portus back up.
A release that fails its checks is rolled back
A new release starts, updates the database and is checked through the same address you use. If it won’t start or a check fails, the previous release goes back into service and the deploy stops.
Every database change is written so the release before it still runs on it, which is what makes the rollback safe.
A restart doesn’t lose a run
Runs are saved step by step. After a restart or a deploy, each run picks up at the step it was on.
A step that fails for a passing reason, like a mail server not answering, is tried again on its own. When it fails for good, the run page says why in plain words and someone can try again from that step. Anything already sent isn’t sent twice.
The run carries on from the step it was on
Security updates go in early on Friday
The server installs the week’s security updates every Friday at 01:00 UTC, 04:00 in Riyadh, when hardly anyone is working. When an update needs a restart, Portus is down for a few minutes, and the status page counts those minutes like any others.
See it run on your own workflow
Bring one workflow your team runs every week. We build it in Portus with you on the call.
Or write toinfo@portusai.co