GitHub for Jira backfill: what it imports, how to restart it, how to check it
On this page
When you connect GitHub to Jira Cloud, the connector does not only link what happens from now on: it also goes back through your repositories and imports past branches, commits and pull requests. In GitHub for Atlassian (the free app many teams still call “GitHub for Jira”) this is called the backfill. This page sums up what Atlassian documents about it, read on 1 November 2026, how to restart it, and how to check that it really brought your history in.
What the backfill imports
From Atlassian’s page Backfill GitHub data in Jira:
- Six months by default. “When you connect GitHub, Jira automatically imports six months of history.” The app’s backfill screen lets you pick an earlier date, and adds that “Branches will be backfilled regardless of their age.”
- Default branch first. On other branches, “only the latest 50 commits from non-default branches will be backfilled.”
- Only what names an issue. Commits are imported only if their message contains an issue key; pull requests need the key in their title, description or source branch name; branches need it in the branch name, the title of their latest pull request or their last commit message.
- Nothing unreachable. “Any unreachable commits (like commits from deleted branches) are excluded.”
So a commit that does not name an issue, or that only lived on a branch you deleted, will never appear, however long the backfill runs. That is the first thing to rule out when history looks incomplete.
How long it takes
Atlassian’s FAQ says that for an organization with little data, backfilling “may only take a few minutes”. For a larger one, the app’s own screen warns that going further back “may take a long time and can slow down other sync work”. GitHub’s API rate limit is the usual reason: a GitHub App gets 5,000 requests an hour as a base, up to 12,500 for large installations, and a backfill has to share them with day-to-day sync.
How to continue or restart a backfill
The documented path, as of 1 November 2026:
- In Jira, go to Apps › Manage apps, then GitHub for Atlassian.
- Open the settings menu (…) of your GitHub organization and choose Continue backfill.
- Pick the date to backfill from, then Backfill data.
- To redo the import from scratch rather than extend it, tick Restart the backfill first. Atlassian describes this as overwriting the existing historical data.
The same screen shows a status per organization: Pending, In progress, Finished or Failed. For a failure, the app says it “will automatically retry”, and offers a retry button next to the status if it does not. It also reports how many repositories could not be completed, and lists the status of each backfill task (pull requests, builds, deployments…) per repository. The causes it names: permission issues, GitHub API rate limiting or network outages.
“Finished” does not prove your history is there
A finished backfill tells you the job ended, not what it brought in. Atlassian’s own tracker has an open request about exactly this: JRACLOUD-99641 (created 25 September 2026, “Gathering Interest”) asks for clear backfill status and sync summary metrics, because the backfill can end with a generic “FINISHED” status even when nothing was ingested.
To check for yourself, in five minutes:
- Pick three issues you know well: one with commits on the default branch, one whose work happened on a feature branch, one older than a few months.
- Count in GitHub: search the repository for the issue key in commits and pull requests.
- Count in Jira: open each issue’s Development panel. Atlassian documents that it shows only the first 100 commits of an issue, so for a long-running issue compare up to 100.
- Count across a project with JQL, for example
project = CHK AND development[commits].all > 0(more queries in JQL for GitHub data in Jira). Run it before and after the backfill: the number of issues should go up, not stay flat. - Explain each gap before blaming the backfill: no key in the message, a deleted branch, more than 50 commits on a non-default branch, or a date older than the backfill start.
What you cannot do today
Three open requests on Atlassian’s public tracker, read on 1 November 2026, describe current limits:
| Request | What it asks for | Status |
|---|---|---|
| JRACLOUD-99984 (28 Oct 2026) | A public REST API to trigger, restart or monitor a backfill. Atlassian’s description: “Currently, there is no public REST API available in Jira Cloud” for it. | Gathering Interest, 0 votes |
| JRACLOUD-94184 (22 Apr 2025) | Re-sync only the affected repositories during a manual backfill | Gathering Interest |
| JRACLOUD-96218 (29 Sep 2025) | Turn the backfill off | Gathering Interest |
In practice: a backfill is started by hand from the settings screen, for a whole organization, and its result is checked by hand.
Does the Forge migration mean a new backfill?
Atlassian is moving GitHub for Atlassian to Forge, with a production rollout in October 2026 and “post-rollout verification” in November and December. Its migration page says your “connected GitHub organizations, repositories, and settings will carry over automatically”. It does not say whether past development data is kept as is or imported again. The safe move is the before/after count above; see GitHub for Atlassian is moving to Forge.
How Linkwright handles history
Linkwright imports less history than GitHub for Atlassian, by design, and says so. On the first read after a repository is attached to a Jira project, and again on every Refresh, it catches up:
- the 100 most recent commits of the default branch;
- the 100 most recently updated pull requests, and the branches;
- the commits of up to 20 branches it has not seen yet (up to 250 per branch); a repository with more such branches catches up 20 more at each Refresh.
It does not go back six months. What it adds is visibility: each repository shows its state (Up to date, Out of date, Webhook down, Rate limited, Failing, Not authorized) with the time of its last successful read, and the Refresh button rereads one repository on demand, without restarting a whole organization. In our side-by-side test on 1 October 2026, GitHub for Atlassian caught up history pushed before connection better at first (3 commits out of 3, against 2 for us); after a fix the same day, Linkwright caught up 3 out of 3 as well.
More guides
- How to connect GitHub to Jira Cloud
- GitHub Issues vs Jira
- GitHub commits not showing in Jira
- Jira Smart Commits with GitHub: syntax and checks
- GitHub for Atlassian is moving to Forge: what to check
- Limit which GitHub repositories each Jira project can link
- JQL for GitHub data in Jira
- Linkwright permissions and sync statuses, explained
- GitHub for Jira alternatives
- 24 hours side by side with GitHub for Atlassian
- GitHub organization not showing in Jira
- GitHub deployments not showing in Jira