aboutsummaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAge
* Use print-backtrace-and-exception/knotsChristopher Baines2025-03-02
|
* Remove the submit-outputs hook from agent-detailsChristopher Baines2025-03-02
| | | | This is now stored and returned by datastore-list-agent-builds.
* Remove unused methodChristopher Baines2025-03-02
| | | | The allocation plan is now in memory.
* Do less work when fetching buildsChristopher Baines2025-03-02
| | | | | | Don't necessarily use a transaction if there are no allocation plan builds, and only run the submit-outputs hook once and store the values, rather than running it each time for each fetch builds request for every allocated build.
* Parallelise part of fetch buildsChristopher Baines2025-03-02
|
* Make the safe simple-format and format variants work with port #fChristopher Baines2025-02-28
|
* Fix the allocation plan build orderingChristopher Baines2025-02-28
| | | | | This was broken in 9d55fe7e17d8a3dd4b0572cb7756c9b33733f084 since sqlite-fold -> sqlite-map reverses the order.
* Try logging about garbage collectionChristopher Baines2025-02-27
|
* Fix unbound with-port-timeoutsChristopher Baines2025-02-24
|
* Drop the timeout for compressing log filesChristopher Baines2025-02-24
| | | | As I'm not sure why this is useful?
* Drop the data service request timeoutChristopher Baines2025-02-24
| | | | As with fibers issue #109, this causes excessive heap usage.
* Move timeouts in to read-drvChristopher Baines2025-02-24
| | | | Rather than the submit-build operation.
* Memoize the read derivationChristopher Baines2025-02-23
|
* Attempt to speed up the allocatorChristopher Baines2025-02-21
| | | | And reduce generated garbage.
* Try a different read-derivation-from-file*Christopher Baines2025-02-20
| | | | | Don't attempt to use the Guix %derivation-cache, just use a fresh hash table each time.
* Better report delays for read-derivation-from-file*Christopher Baines2025-02-20
|
* Remove with-fluids in read-derivation-from-file*Christopher Baines2025-02-20
| | | | Since this should be set for the entire service.
* Increase the size of the utility thread poolChristopher Baines2025-02-20
|
* Better instrument the utility thread poolChristopher Baines2025-02-20
|
* Better count the in use parallel hook thread pool threadsChristopher Baines2025-02-20
|
* Reduce the number of threads in the reader poolChristopher Baines2025-02-20
| | | | As I think up to 32 is unnecessary.
* Add more delay logging for substituting and reading derivationsChristopher Baines2025-02-20
|
* Remove the thread metricChristopher Baines2025-02-19
| | | | This duplicates a metric from the process metric updater.
* Instrument hook threads being usedChristopher Baines2025-02-19
|
* Remove knots workaroundChristopher Baines2025-02-19
|
* Log the relay from reading derivationsChristopher Baines2025-02-19
|
* Instrument the reader thread poolChristopher Baines2025-02-19
|
* Replace the builds_unprocessed indexChristopher Baines2025-02-19
| | | | | With builds_live, for all unprocessed and non canceled builds. This should be more useful, especially for queries to cancel builds.
* Guard against negative state ids being requestedChristopher Baines2025-02-18
|
* Prioritise and instrument the get state transactionChristopher Baines2025-02-16
|
* Use default priority for datastore-update-unprocessed-builds-for-build-successChristopher Baines2025-02-16
| | | | As it's still quite important.
* Instrument the writer queuesChristopher Baines2025-02-16
|
* Remove old data correction codeChristopher Baines2025-02-16
|
* Use three priority levels rather than twoChristopher Baines2025-02-16
| | | | To try and make better use of the database.
* When fixing all_inputs_built, look at the most recent builds firstChristopher Baines2025-02-16
|
* Always call metric-observe-duration in datastore-call-with-transactionChristopher Baines2025-02-16
| | | | Regardless of the duration.
* Support key only tag queries where there are lots of valuesChristopher Baines2025-02-16
| | | | Previously the code would generate a query which was too large.
* Support canceling builds by created_atChristopher Baines2025-02-16
|
* Stop printing reader procsChristopher Baines2025-02-16
| | | | Unless every slot is currently used.
* Fix datastore operation nestingChristopher Baines2025-02-16
| | | | | This was dependent on the overlapping argument paramters for the thread pools, which changed with the use of the knots thread pool.
* Use safe output procedures in the sqlite moduleChristopher Baines2025-02-16
|
* Add format/safeChristopher Baines2025-02-16
|
* Fix listen repl log messageChristopher Baines2025-02-16
| | | | And don't enable the repl by default.
* Add some code to check and correct the unprocessed builds tableChristopher Baines2025-02-15
| | | | | I think the recent changes to build-success handling caused some updates to be missed.
* Add a listen replChristopher Baines2025-02-15
| | | | To make it eaiser to access the running service.
* Remove duplicate workaroundChristopher Baines2025-02-15
| | | | This is done in perform-coordinator-service-startup.
* Move installing suspendable ports to run-coordinator-serviceChristopher Baines2025-02-15
| | | | To make the behaviour more similar when using the script vs the guix service.
* Have the returned allocation plan filtering be optionalChristopher Baines2025-02-15
| | | | So that this better reflects the plan.
* Move creating the pid fileChristopher Baines2025-02-13
|
* Replace with-throw-handler with with-exception-handlerChristopher Baines2025-02-13
|
generated by cgit v1.2.3 (git 2.47.1) at 2025-03-19 05:18:43 +0000