aboutsummaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAge
...
* Log delays for the coordinator thread channelsChristopher Baines2021-12-21
|
* Log delays for procedures involved in handling build resultsChristopher Baines2021-12-20
|
* Log delays in fetch builds to allocateChristopher Baines2021-12-20
|
* Log delays when worker thread calls occur inside worker threadsChristopher Baines2021-12-20
|
* Reduce the time in the transaction when replacing the allocation planChristopher Baines2021-12-20
|
* Log delay from select allocated buildsChristopher Baines2021-12-20
|
* Log delays for entering transactionsChristopher Baines2021-12-20
|
* Increase precision of delay loggingChristopher Baines2021-12-20
|
* Fix displaying the total bytes for uploads in progressChristopher Baines2021-12-19
|
* Fix logging deleting temporary filesChristopher Baines2021-12-19
|
* Cleanup temporary files on startupChristopher Baines2021-12-19
|
* Compress outputs in the thread for the buildChristopher Baines2021-12-19
| | | | | Since this can take a significant amount of CPU, and doing this upfront helps when there's lots of outputs to upload.
* Change the post build actions thread pool sizeChristopher Baines2021-12-19
| | | | | To be that of the number of parallel uploads, since that is what this now controls.
* Report the total bytes when reporting upload progressChristopher Baines2021-12-19
|
* Stop using with-upload-slot for queuing uploadsChristopher Baines2021-12-19
| | | | As that's handled by the thread pool instead.
* Increase the number of chunked request worker threadsChristopher Baines2021-12-19
| | | | | | | Since I think 8 might not be sufficient with lots of agents. Ideally the issue preventing the fibers threads from reading the requests would be resolved, this is just a workaround.
* Don't use an unlimited number of threads for post build actionsChristopher Baines2021-12-19
| | | | | | Otherwise, when there's a large queue of outputs to upload, there will be a large number of threads, which can cause issues like exhausting file descriptors.
* Don't try to exit when hook threads crashChristopher Baines2021-11-29
| | | | | | | It doesn't seem to work, just wait a bit and try again. I've seen a few cases of "conversion to port encoding failed" around here, but I'm not sure exactly where they're coming from.
* Check before deleting filesChristopher Baines2021-11-26
| | | | As I've seen exceptions here.
* Remove redundant if in the controllerChristopher Baines2021-11-26
|
* Don't print backtraces in the controller when chunked inputs endChristopher Baines2021-11-26
|
* Don't log backtraces for chunked input issues in worker threadsChristopher Baines2021-11-26
| | | | As it's something that can be expected.
* Use a specific exception type for chunked input issuesChristopher Baines2021-11-26
| | | | So that it can be handled specifically.
* Allow disabling the logging of exceptions in worker threadsChristopher Baines2021-11-26
| | | | | As some exceptions can be expected, like chunked responses being incomplete, so allow reducing the unnecessary output.
* Drop parallel uploads to 1 againChristopher Baines2021-11-26
| | | | | | I've seen agents crashing with the "Too many heap sections: Increase MAXHINCR or MAX_HEAP_SECTS" error, hopefully reducing the parallel uploads will help avoid this.
* Delete existing files when processing upload requestsChristopher Baines2021-11-22
| | | | I think this will help when handling new requests after failed ones.
* Unwind on some exceptionsChristopher Baines2021-11-22
| | | | The error handling here should be handling by unwinding.
* Improve some way numbers are displayedChristopher Baines2021-11-22
|
* Only check the size of the file once when uploadingChristopher Baines2021-11-21
|
* Fix variable reference in submit-outputChristopher Baines2021-11-21
|
* Log delays with datastore-insert-buildChristopher Baines2021-11-20
|
* Improve delay logging in datastore-call-with-transactionChristopher Baines2021-11-20
| | | | Make sure to log the delay in the calling thread.
* Log delays in insert-derivation-and-return-outputsChristopher Baines2021-11-20
| | | | As I've seen this be quite slow.
* Use prepared statements when inserting derivationsChristopher Baines2021-11-20
| | | | As I hope this might be a little faster, plus it's also neater.
* Always reset the statement in db-output->output-idChristopher Baines2021-11-20
|
* Compress outputs outside of the upload slotChristopher Baines2021-11-20
| | | | | So that the only thing taking place in the upload slot, is the actual upload, which should improve throughput.
* Add an index on output_metadataChristopher Baines2021-11-18
| | | | This should speed up fetching builds.
* Monitor a few more sqlite related proceduresChristopher Baines2021-11-18
|
* Improve the datastore-call-with-transaction delay loggingChristopher Baines2021-11-18
| | | | Associate the delay with the passed proc.
* Improve the call-with-delay-logging procedureChristopher Baines2021-11-18
| | | | | Support arguments to the passed procedure, and report more of the things involved in the monitored time period.
* Improve delay logging seconds reportingChristopher Baines2021-11-18
|
* Add delay logging to fetching buildsChristopher Baines2021-11-17
| | | | As this appears to be a bit too slow.
* Track delays for reporting metricsChristopher Baines2021-11-16
|
* Track delays for a couple of sqlite datastore operationsChristopher Baines2021-11-16
|
* Add a tracing style delay loggerChristopher Baines2021-11-16
| | | | This tracks delays and can report a breakdown by procedure.
* Track worker thread task durationsChristopher Baines2021-11-16
| | | | So this can be logged.
* Remove TODO from the README about strings as foreign keysChristopher Baines2021-11-16
| | | | This has pretty much been done.
* Check if an output has been uploaded before trying to upload itChristopher Baines2021-11-16
| | | | | | | | | | This can help if the output has been uploaded, but the hash isn't present, since trying to submit the build result will prompt for the output to be sent again, but it doesn't need to be, the agent just needs to wait. This is a little inelegant, maybe there needs to be some way for the agent to explicitly check for the hash to be computed, but I'm hoping these changes will help with uploading large outputs.
* Move the timing of fetch-builds around more of the codeChristopher Baines2021-11-16
| | | | As it was just covering part of the action.
* Add error handling around computing output hashesChristopher Baines2021-11-15
| | | | As I've seen decompression errors.
generated by cgit v1.2.3 (git 2.25.4) at 2024-10-03 01:56:01 +0000