Skip to content

Write Queue

Write Queue shows the changes Colony has queued to write to your VCS, and the ones that did not land. Route: /operations/throughput?view=writes

Colony sends most writes to your VCS directly. When a write fails with a temporary error, such as a rate limit or an outage, Colony queues it and retries it until it lands or runs out of attempts. A few writes are queued from the start. When a PR, comment or label you expected never appeared, check here first: this page shows the writes that are waiting or failed, and where you retry or discard them. Not every write can appear here. Merges made by the Merger agent, reviews of external pull requests, and writes refused with a permanent error have no row.

The page follows the tenant chosen in the sidebar tenant selector. Switching tenants reloads the list from the first page.

The Filter by status control picks which writes are listed. It opens on Failed.

StatusMeaningEmpty-state message
FailedThe write used its attempts and did not land. These need you.”No failed writes. Your VCS is in sync.”
PendingQueued and waiting to be sent, or waiting for its next retry.”Nothing waiting. Every write has been sent to your VCS.”
ProcessingBeing sent to your VCS right now.”Nothing in flight right now.”
CompleteLanded in your VCS.”No completed writes to show yet.”
DiscardedDropped by a person. It was not retried and the change was not made.”No discarded writes.”

The filter is kept in the ?status= query parameter, so /operations/throughput?view=writes&status=pending is a linkable view. An unknown value falls back to Failed.

ColumnWhat it shows
RepositoryThe owner/name the write targets. Links to the Ledger filtered to that repository (/work?repo=owner/name).
IssueThe issue number (#123), or — when the write isn’t tied to an issue.
ActionThe write type.
StatusThe write’s current status.
AttemptsAttempts used out of the maximum, as attempts/maxAttempts.
ErrorThe last error, truncated. Hover for the full text. — when there is none.
UpdatedHow long ago the row last changed.
ActionsRetry and Discard, where they apply.

Retry appears only on failed rows. It resets the attempt count to 0 and puts the write back in the queue as pending, so it is sent again straight away.

Discard appears on failed and pending rows. It asks you to confirm first. A discarded write is not retried and the change is not made in your VCS. Choose Discard write to confirm or Keep write to leave it alone.

If a row changed underneath you (for example, it was retried by a teammate or landed while the page was open), the action is refused and the list refreshes to show the real status.

The list shows 50 writes per page, newest first, with Previous and Next controls and a “Showing x–y of total” line. The controls appear only when there is more than one page.

These endpoints take a signed-in dashboard session, not a connection token.

EndpointPurpose
GET /pipeline/projections?status=&limit=&offset=&tenant=Lists writes. status defaults to failed and an invalid value returns 400. limit defaults to 100 (maximum 500).
POST /pipeline/projections/:id/retry?tenant=Re-queues a failed write.
POST /pipeline/projections/:id/discard?tenant=Discards a failed or pending write.

Retry and Discard return { "ok": true }. A malformed id returns 400, an unknown write returns 404, and a write whose status doesn’t allow the action returns 409.

Both pages collect things that went wrong, but they track different failures.

  • Dead Letters (/operations/throughput?view=failed) holds failed pipeline transitions: an issue that could not move to its next state. See Recover a Dead-Lettered Issue.
  • Write Queue holds VCS writes that did not land: the pipeline moved on, but a comment, label or queued merge never reached your VCS.

If an issue is stuck, start with Dead Letters. If the pipeline looks fine but your repository is missing a change, start here.