You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Added
queue.disk keeps the queue of a push output on disk (ADR 0015): samples already paid for in API quota survive a backend outage longer than the memory queue and a restart of Tapyr, within maxSize (default 512MiB) and maxAge (default 2h). tapyr plan says how long an outage each buffer holds; the Outputs page, the status API, the tapyr_output_buffer_* self-metrics, the Tapyr / Overview dashboard and the TapyrOutputBehind and TapyrOutputBufferNotWritable alerts show it.
Fixed
The relative times of the web UI ("next run in 45s", "2m ago", "resets in 2h", the uptime) stayed frozen between two runs; they move every second, and a next run whose time has passed reads "due".
A reload that removed or changed a push output whose backend did not answer waited for it forever; the flush of the outputs a reload stops is bounded to 10 seconds, as at shutdown.
tapyr plan, the plan problems and the logs named a quota scope keyed by a credential by its hash ("credential 34829e408ebd22e8"); they name it by the accounts that share the credential, as the Budgets page does.
tapyr validate answered interval: 5 with a hint to quote the value, which led to the next error; a duration, size, rate or percentage written as a bare number now says what the value looks like (30s, 5m or 1h30m).
The Graph cut the longest job names of the built-in packages (meraki-appliance-uplink-statuses) and the upper-case "fan-out of" line of their tiles; the tiles are wider and the fan-out line is in lower case.