First 60 minutes when Odoo production is down
Direct answer: Freeze deploys, snapshot the database and the filestore, read the traceback, and roll back the last change if the failure started there. Reproduce custom-module errors on staging. Do not run -u on production while the registry is broken.
- Name the severity before anyone edits data. P1 means the business cannot invoice, ship, receive, or post. P2 means one of those is blocked for a department, or a storefront, EDI, or 3PL feed has stopped.
- Freeze deploys. Tell the team not to click Upgrade on the Apps screen. A second -u on a broken registry makes the rollback harder.
- Pick one internal owner. That person tells the warehouse and finance what is blocked and when the next update is. Everyone else stays out of the database.
- Take a pg_dump (or a host snapshot) and copy the filestore to storage this server cannot overwrite. Do that before a restart, a module update, or a SQL edit.
- Read the Odoo log from the minute the failure started. Copy the traceback. Note the last deploy, the last module update, and the last app installed.
- Check free disk, PostgreSQL connections, and locks in pg_stat_activity. If the volume is full, free space or expand it before you restart Odoo. If one backend holds the lock, record its query, then cancel that backend.
- If the failure started at the last deploy or module update, roll that change back from the snapshot. Confirm a user can log in, validate one delivery, and open one invoice before the floor comes back in.
- Open scheduled actions and look for jobs left in progress. If invoices or delivery notes are not leaving the system, open the mail queue and the outgoing mail server next.
- If the traceback names a custom module, copy production to a neutralized staging database. Reproduce with odoo-bin -u that_module -d staging --stop-after-init. Do not retry -u on production.
- Take the failing custom modules off the production addons path only after the snapshot is stored and staging has reproduced the error. Restart, and confirm the registry loads.
- After writes work again, refresh an expired integration token and repair the outgoing mail server. Replay failed jobs from the queue. Do not retype orders by hand.
- Write the timeline: what changed, what you restored, and what is still degraded. Schedule the real fix in a window. Leave production on the restored system until staging passes.
What makes an Odoo production outage worse?
Direct answer: A second upgrade, a restart with no snapshot, and a hand-typed replay of integration orders all turn a rollback into a data repair.
- Running -u on production while the registry traceback is still open.
- Restarting Odoo before the database and the filestore are copied.
- Giving every user Settings to get past an Access Error.
- Editing scheduled actions or invoice sequences in SQL on the live database with no snapshot.
- Replaying storefront, EDI, or 3PL orders by hand and creating duplicates.
- Treating Odoo.sh hosting support, or Odoo's own functional support, as a fix for your custom modules.
Which Odoo versions does this outage playbook cover?
Direct answer: The same failures show up on Odoo 16, 17, 18, and 19, Community and Enterprise, on Odoo.sh and on your own servers. Odoo 14 and 15 still show up in repair work.
Use this playbook on Odoo 16, 17, 18, and 19, Community and Enterprise, whether you host on Odoo.sh or your own servers. Odoo 14 and 15 still appear when a series change is what took production down. Stock reservation changed in 17, and views changed again in 18, which is why a custom module that loaded yesterday can stop the registry today. Odoo's own support covers standard apps. Custom modules, cron jobs, and integrations stay with your team or your firm. The ERP Firm plans are $5,000/mo up to 15 users, $8,500/mo for 16–50, and $14,000/mo for 51–150.
What does emergency Odoo help cost, and who owns the code?
Direct answer: These figures are planning estimates, not client statistics, for an hourly emergency partner, a freelancer marketplace, a full-time hire, Odoo official support, and a flat monthly team.
Planning estimates, not client statistics. Hourly totals are the published rate times hours. The salary figure is a planning cost. None of these are measured client averages.
| Option | Response time | Who owns custom code | Monthly cost range | Upgrade safety |
| Hourly emergency partner | Often a multi-hour minimum, then next-business-day response | Often the firm's repository, unless the contract assigns the code to you | $175–$250/hr, open-ended. A 10-hour incident is $1,750–$2,500. Not a monthly fee. | Custom modules are retested only if you buy that time |
| Freelancer marketplace | Hours to days if someone bids. No written response time | Often one laptop until you require a push to your Git | Hourly, often with no cap and no monthly fee | Low. Context leaves when the contract ends |
| Full-time hire | About 90 days to hire. No coverage on the day the system goes down | Your Git, after you create the repository | About $165,000/yr loaded. Not a monthly fee. | One employee. Vacation and resignation reopen the gap |
| Odoo official support | The queue on the subscription you buy from Odoo. Functional questions only. Not a custom-code emergency desk | Odoo does not take your custom modules, and standard support does not fix them | Included with Enterprise fees paid to Odoo. Community has no vendor support agreement | Standard apps and Studio. Custom modules stay out of scope unless you hold a maintenance of customizations subscription |
| Flat monthly team (The ERP Firm) | 1-hour first response, 9am–6pm PT on business days. 24/7 emergency support for outages that stop the business | Your Git repository | $5,000/mo up to 15 users. $8,500/mo for 16–50. $14,000/mo for 51–150 | One version upgrade a year is included and rehearsed on staging. Unhappy in the first 30 days? Cancel with a full refund. Ongoing care is Odoo Autopilot. |
Severity triage for an Odoo outage
Direct answer: Pick the row that matches the blocked job, then use the target window to decide who to call. These are planning targets, not measured client response times.
Planning estimates, not client statistics. Response and fix windows are triage targets for a senior engineer, not a measured client average.
| Severity | Example Odoo symptoms | Target response | Target fix window |
| P1 production down | Users cannot log in, or they cannot confirm a sales order, validate a delivery, receive stock, or post an invoice. Workers crash, or PostgreSQL refuses connections. | 15 minutes | Restore service in 1–4 hours |
| P2 major degraded | One department is blocked while the rest of Odoo loads. Stock on the floor does not match Odoo. A storefront, EDI, or 3PL feed has stopped. | 1 hour | Same business day |
| P3 workaround exists | Staff can finish the job with a spreadsheet, an export, or one admin clicking through. The error repeats, and today's shipments still leave. | 4 business hours | 1–3 business days |
| P4 cosmetic | A report label, a view glitch, or a warning that does not block a document. | Next business day | Next planned release |
Ten Odoo outage causes
Direct answer: Start with the last deploy and the traceback. Registry errors, a full disk, and a stuck cron are faster to undo than a sequence or a record rule, which need finance or an access review.
Planning estimates, not client statistics. Time to fix is recovery time for one senior engineer, not a measured client average.
| Symptom | Root cause | Fix | Typical time to fix |
| Module update fails and the registry will not load | odoo-bin -u, or the Apps upgrade button, hit a view xpath, a missing dependency, or an inconsistent module state | Restore the pre-upgrade database if users are blocked. On a copy, fix the module and rerun -u --stop-after-init | 2–8 hours if the rollback is clean. 1–3 days if the registry has to be repaired on a copy |
| Traceback after an Odoo series change | A custom addon still calls an API removed between series, such as stock reservation in 17 or a view that moved in 18 | Take that addon off the production addons path after a snapshot. Port it on a neutralized staging database | 4–16 hours to restore service. The port is a separate change |
| 502s, workers recycling, or a stuck bus | Worker count times the hard memory limit exceeds RAM, or the longpolling (gevent) port is not proxied | Size workers and limit_memory_hard to the host. Proxy the bus port. Restart only after the database snapshot | 1–4 hours |
| Requests hang, or the disk is full | A session holds a lock, a DDL statement wants an exclusive lock, or the volume for PostgreSQL or the filestore is full | Record the query in pg_stat_activity, then cancel that backend. Free disk or expand the volume before you restart Odoo | 1–4 hours |
| Scheduled jobs stuck on Running | An ir.cron worker died and left the job in progress, or the job runs longer than its interval | Clear the stuck scheduled action from the technical menu on a snapshot first. Fix the job, then shorten the interval | 1–6 hours |
| Invoices and deliveries never email | The outgoing mail server rejects authentication, or the mail queue cron is off, so mail.mail rows sit in exception | Repair the outgoing server, retry or cancel the failed messages, and confirm the mail queue cron | 1–4 hours |
| Storefront, EDI, or 3PL jobs return 401 or 429 | The OAuth token expired, or retries with no backoff hit the vendor rate limit | Pause the queue, put a fresh token on the integration user, and replay the dead-letter rows after backoff | 2–8 hours |
| PDFs and images 404 | The database was restored without the matching filestore, or data_dir now points at an empty path | Restore the filestore that belongs to that database dump. Do not recreate files by re-saving records | 2–8 hours when a matching backup exists |
| Cannot post, or invoice numbers skip | A company lock date rejects the move, or the invoice sequence moved during an upgrade | Confirm the lock date with finance before you change it. Repair the sequence on a copy, then post only the blocked moves | 2–8 hours, plus finance sign-off |
| Access Error after an upgrade, or empty lists | The upgrade reset groups on res.users or rewrote ir.rule domains | Compare groups and record rules with the pre-upgrade database and restore the intended groups. Do not give every user Settings | 2–8 hours |
Prevention checklist before the next outage
Direct answer: A backup you have not restored, and an upgrade you have not rehearsed, are the two gaps that turn a bad deploy into a long outage.
Planning checks, not client statistics. Prove each row on a calendar, not after the next outage.
When should you call an Odoo emergency support team?
Direct answer: Call when the database is already down and an hourly clock would start before anyone reads the log. The ERP Firm is $5,000/mo up to 15 users, $8,500/mo for 16–50, and $14,000/mo for 51–150. First response is within 1 hour, 9am–6pm PT on business days. Outages that stop the business go to 24/7 emergency support answered by a senior engineer. Unhappy in the first 30 days? Cancel with a full refund.
If the work involves custom modules, an upgrade, a storefront or warehouse integration, or a production outage, a focused engagement is usually faster than recruiting. See the related service.
See our industry practice
Related Odoo guides
More answers in the Support cluster and nearby topics:
Commercial pages