Who runs ERPNext after go-live? Someone has to apply its updates, keep backups that restore, patch security holes, watch the server and answer the phone when it breaks. You can do that yourself on a server you rent, pay Frappe Cloud to run the platform, or pay a managed host to run the lot. This guide lists the work, shows who does each job on each path, and ends with a chooser that can say: do it yourself.
An ERPNext that nobody owns runs fine until the night it does not. I lead ERPNext and Frappe programmes at MithTech, a team of Frappe specialists in Bengaluru. We sell managed hosting, so one of the three answers pays us. Nothing pays us for hosting if you pick Frappe Cloud or run it yourself, and for some teams that is the right pick.
Short answer
Run it yourself if a named person with a deputy already owns servers and a restore has been tested. Choose Frappe Cloud if you want the platform run for you and your ERPNext is close to standard. Choose a managed host if one party must answer, in writing, for upgrades, your own apps and the night call. With nobody who runs servers, start with the second or third.
What work is left after go-live?
Eight jobs continue after go-live on every path: applying ERPNext's point releases, applying security fixes, taking backups, proving a backup restores, moving to a new major version, watching the system, answering when it breaks, and helping users with how ERPNext works. The path you choose decides who does each job. It does not remove one.
Four numbers show the pace. The first two were counted on 13 October 2026 from GitHub's own lists.
- Point releases. From 15 July to 13 October 2026, the ERPNext repository published 36 stable releases: 18 on version 15 and 18 on version 16. That is one on each line about every five days (ERPNext releases).
- Security advisories. In the same 91 days,
frappe/erpnextpublished 75 advisories andfrappe/frappepublished 20. Thirty-four of the 95 are rated high or critical, and every one names the release that fixes it (ERPNext advisories, Frappe advisories). A long list partly means people are looking; what matters is how soon a fix reaches your server. - Support dates. Support for version 14 ended on 31 January 2026. Frappe plans version 15 to the end of 2027 and version 16 to the end of 2029 (Frappe support table).
- A major upgrade is a project. Version 16 needs Python 3.14, Node 24 and MariaDB 11.8 (install requirements). Its migration notes list six breaking changes, two of them in tax data (v16 migration notes).
Who does each job on each path?
On your own server you do every job. On Frappe Cloud, Frappe does the platform jobs, and on a public bench the updates too; you keep major upgrades and your own apps. With a managed host the contract decides, so the last column says what to ask for. The Frappe Cloud column comes from Frappe's own pages and Frappe Cloud's source code, read on 13 October 2026.
A bench is the set of apps, at set versions, that sites run on. Frappe maintains a public bench, and it takes no apps of your own. A private bench is under your control (Frappe Cloud benches page).
Keeping it current
| Job | Run it yourself | Frappe Cloud | A managed host |
|---|---|---|---|
| ERPNext point releases | You pull and apply them. bench update backs up first unless you pass --no-backup. | Public bench: Frappe deploys the updates, and the site's auto-update switch moves it in off hours. Private bench: you deploy the new bench version, then auto-update moves your sites. | Tried on a copy, then applied on a schedule agreed with you. |
| Security fixes | Yours the day the advisory is published. Nothing pushes them to you. | Public bench: Frappe applies the patches, and auto-update moves the site. Private bench: you choose which apps to update, so a fix lands when you deploy it. Frappe's page does not say it does that for you. | Applied on the schedule in the agreement. Ask how soon after publication. |
| Major upgrade, 15 to 16 | A project: newer Python, Node and MariaDB, rehearsed on a copy. | You start it from the dashboard (upgrade dialog). Public bench: the site is upgraded to the next version. Private bench: you need a private bench on the new version holding every app the site uses, which the dashboard can deploy if each app has a branch for it. Either way it will not skip a major version. | Rehearsed on a copy with your custom apps. You check the copy before the date is set. |
| Your own apps | Yours to test and fix. | They need a private bench. Testing them against each version is yours. | In scope only if the contract says so. |
Keeping it safe and answered
| Run it yourself | Frappe Cloud | A managed host | |
|---|---|---|---|
| Backups | A bench takes one every 6 hours and keeps three, on the same server. Copying them off is your job. | Every 24 hours. Offsite copies kept 7 daily, 4 weekly, 12 monthly and 10 yearly, on plans of USD 25 and above. | Copies off the server, with the count and the retention written down. |
| Proof a backup restores | A drill you schedule. bench restore needs at least the database file. | Frappe's backup docs describe no restore test. | A restore onto a separate copy on a schedule, with a note saying which and when. |
| Watching it, and the night call | You, with an outside uptime check, disk and queue alerts and a written call list. | Monitoring is Frappe's. Response terms depend on the plan, and the pricing page states no numbers. | An on-call engineer and response times by severity. Get both in writing. |
| How ERPNext works for your business | Your team, your implementer or the forum. | The warranty covers bug fixes and "does not include functional support". | Rarely part of hosting. That is the support line, usually an AMC. |
| Where the data sits | Wherever you rent the server. | Your choice of region; Mumbai is one. | Your cloud account or theirs. Ask for the region in writing. |
| Getting out | Nothing to get out of. | Offsite backups can be downloaded. Frappe's docs describe moving a site in, not out. | Ask what you get, in what format and how fast. |
What does running it yourself involve?
Running ERPNext yourself means one named person owns seven jobs: the route, the backups, the restore test, the updates, the security fixes, the major upgrades and the watch on the server. It costs the server's rent plus that person's time, and gives you full access to the bench.
Choose the route
Frappe's production-setup page warns against bare-metal installs and recommends Docker images (setup production). The official container setup is frappe_docker; a bench on a plain VM is the other route. Know which you run, because the update and backup steps differ.
Set up backups, then move them off the server
bench init adds a cron job that backs up all sites every six hours (bench source). Frappe then deletes the oldest sets beyond three by default (Frappe source). Three backups six hours apart cover less than a day, on the same disk as the site. Push them to object storage; frappe_docker has a script for that, and its docs say restores are manual (frappe_docker operations).
Restore one before you need it
bench restore needs at least the database file, cannot restore files alone and does not support downgrades (restore reference). Restore last night's backup onto a fresh server and open an invoice. One Frappe forum user got a success email while the S3 bucket stayed empty (forum thread, 2021). Check the bucket, not the email.
Apply point releases on a copy first
bench update pulls the apps, applies patches, builds assets and migrates (bench commands). On a small server the asset build can run out of memory (the fix ladder). On Docker, a patch upgrade from 15.16.2 to 15.17.0 has left a site's CSS and JavaScript broken (frappe_docker issue 1353). Both are cheaper to meet on a copy.
Apply security fixes in the week they are published
Subscribe to the releases of both repositories. Each advisory names the release that fixes it, so the job is to know your version and move.
Plan the major upgrade as a project
Rehearse it on a copy that has your custom apps, and move one major version at a time. Frappe Cloud's own upgrade code refuses to skip one (source); do the same on your server. Back up first, because Frappe does not support downgrading a site. The latest-version guide has the dates.
Watch it, and write down the phone
Add an uptime check from outside your network, disk-space and queue alerts, and a written name for who is called, with a deputy. A server only its builder understands fails when the builder is on leave.
It is the right answer when a named person already runs servers and has a deputy, a restore has been tested, and either a day without ERPNext is survivable or that person can be on call. Heavy custom code that needs full access to the bench also points here. If that is you, run it yourself; we would rather you run it well than buy hosting you do not need.
What does Frappe Cloud take off your hands, and what stays yours?
Frappe Cloud is Frappe's own hosting platform. Frappe says it takes care of installation, setup, upgrades, monitoring, maintenance and support (frappe.io/cloud). In practice the platform jobs move to Frappe and the decisions stay with you.
What it takes off your hands:
- Servers and monitoring. The servers run on cloud providers such as AWS and OCI, with Mumbai among the regions. Monitoring is part of the platform.
- Backups. Every 24 hours, with offsite copies in a Mumbai S3 bucket by default, on plans of USD 25 and above (backup docs).
- Updates and security patches, on a public bench. Frappe deploys them. Each site has an auto-update switch, and when it is on the site moves in its off hours (Frappe Cloud source).
What stays with you:
- Major upgrades. You start them. A forum user who hit the private-bench error was blocked until a private bench on version 16 held every app the site used (forum thread).
- Your own apps, and the private bench they need. A public bench takes no custom apps. On a private bench you choose which apps to update and deploy the new bench version, and auto-update then moves your sites (deploying a bench). Testing your apps against each version is yours.
- Functional help and support terms. The Product Warranty covers bug fixes in Frappe apps and Frappe Cloud and "does not include functional support". Priority support is optional on the Servers plan; the pricing page states no response times or uptime.
- Restore proof and the exit. The backup docs describe no restore test, and the migration docs cover moving a site in, not out (migrating a site).
Frappe lists Sites from $5 (₹410) a month and Servers, whole virtual machines, from $40 (₹3,600) a month, with a 14-day free trial. Both are starting prices (Frappe Cloud pricing, read 13 October 2026). The same page says Hetzner suits development and testing, and recommends dedicated AWS or OCI instances where uptime and durability are critical.
It is the right answer when nobody on your side runs servers, your ERPNext is close to standard, and you can live with Frappe's support terms. Functional questions still need your implementer or a support contract. Application support is a separate line wherever the site runs; ours is on Support & AMC.
What does a managed host add, and what should it put in writing?
A managed host sells accountability: one party answers for the server, the backups, the upgrades and the night call, under a written scope. What you pay for is the written part, so read the scope before the price. Ask any provider for six things.
- Who holds the cloud account and root. Your account or theirs, and what you get on request.
- How backups work. How often, where, how long kept, and a restore onto a separate copy on a schedule, with a note saying which.
- How upgrades work. Point releases and major versions, tried on a copy with your own apps, and checked by you before the date is set.
- Response and resolution times by severity, and the hours they run.
- What sits outside the scope. New modules, large integrations and help with how ERPNext works usually do.
- The way out. What you get, in what format, in how long.
MithTech's managed hosting is described on its own page: your cloud account or ours, backups restored to prove them, upgrades tried on a copy first and an exit kit when it runs on our servers. It is priced after a call, in writing, by apps and sites, backup and monitoring level and response time, with no per-user charge. Response and resolution targets are on Support & AMC; the rest of the project is priced by stage in the stage calculator.
Which path fits your team?
The chooser weighs who runs servers most, then downtime, then custom code, then your version habit.
Answer four questions about your team, your downtime, your version habit and your custom code to see run it yourself, Frappe Cloud and a managed host ranked, with the reasons, what you carry and a first test.
Four answers, three paths
Who on your side runs servers?
If ERPNext is down for half a day, what happens?
How will you handle new ERPNext versions?
How much of ERPNext have you changed in code?
Answer all four to see the three paths.
We tried all 81 combinations. Running it yourself ranks first in 24, Frappe Cloud in 30 and a managed host in 27. When two paths are close, run the first test for each and drop the one you cannot pass.
What happens if nobody owns it?
An ERPNext that nobody owns keeps running until a fix is missed. It fails in a sequence: backups stop being copied, updates stop, a security fix is published and not applied, the version passes its end of support, and then one outage needs a restore nobody has tried.
Version 14 shows where that ends: no Frappe Support and no latest bug fixes since 31 January 2026 (Frappe forum).
Where does this leave you?
| Run it yourself | Frappe Cloud | A managed host | |
|---|---|---|---|
| Choose it when | A named person with a deputy owns servers and a restore has been tested | You want the platform run for you and your ERPNext is close to standard | One party must answer for upgrades, your own apps and the night call, in writing |
| You carry | Patches, backups, restores, upgrades, monitoring, the phone | Major upgrades, your own apps with a private bench to deploy them, functional questions | Choosing well, the upgrade date, reading the monthly note |
| First test | Restore last night's backup onto a fresh server this week | Run a month-end on the free trial with a copy of your site | Ask for the scope, the response table and the exit before you ask the price |
| Move off it when | The person leaves, or ERPNext turns month-end critical | Your response needs outgrow its support terms, or a contract asks for infrastructure in your own name | You hire a person with a deputy, or the scope stops matching the bill |
We earn a hosting fee only when a team picks the third column. That is why the first test for each path is something you can run without us.
Frequently asked questions
+Who maintains ERPNext after go-live?
Whoever you choose; the job does not disappear. On your own server it is your team or your implementer. On Frappe Cloud, Frappe maintains the platform and you answer for major upgrades and your own apps. A managed host maintains the stack under a written scope. Write down a name and a deputy.
+Is Frappe Cloud better than self-hosting for production?
Better for a team with nobody who runs servers and an ERPNext close to standard. Not better for a team with such a person who wants full access to the bench. Frappe's pricing page recommends dedicated AWS or OCI servers for mission-critical workloads, so read your plan's support terms.
+Does Frappe Cloud upgrade ERPNext automatically?
Point updates can be automatic, depending on the bench. On a public bench Frappe deploys updates and security patches, and each site's auto-update switch moves it in off hours. On a private bench, which your own apps need, you deploy the new version, and auto-update then moves your sites. Major upgrades are not automatic: you start them, and Frappe Cloud refuses to skip a major version.
+Can I move from Frappe Cloud to my own server later?
Frappe's docs describe moving a site in, not out. You can download an offsite backup from the dashboard and restore it on your own bench at the same version. Try that on a copy before you need it.
+What happens if we never update ERPNext?
It keeps working until something forces the issue: a fix was never applied, or the version reaches the end of support. Version 14 reached it on 31 January 2026; version 15 is planned to reach it at the close of 2027.
+Do I need a managed host if I already have an IT person?
Not necessarily. Ask what happens at night, on leave and when that person leaves. If a deputy covers those and a restore has been tested, running it yourself is sound. Buy only what the person cannot cover, such as the night call or major upgrades.
+Which ERPNext version should I run?
Version 16 for a new go-live. It was released on 12 January 2026, and Frappe plans to support it to the end of 2029. Version 15 is planned to the end of 2027. Version 14 is out of support.
What is the bottom line?
Someone owns ERPNext after go-live, by name or by default. Pick the owner while the system is healthy, test a restore, and write the answer down. All three paths are sound for the right team, and the wrong one is the path nobody chose.
Comments & ratings
No comments yet. Start a new discussion.
Manoj · Published on 14 October 2026
Worked with us? Leave a Google review