| Commit message (Collapse) | Author | Age | |
|---|---|---|---|
| * | Try to prioritise some database operationsHEADmaster | Christopher Baines | 2024-10-14 |
| | | | | | | | At the moment, hooks can get stuck behind new builds being submitted due to the slow writes. Try to manage this by prioritising the write operations from the hooks. | ||
| * | Add make-discrete-priority-queueing-channels | Christopher Baines | 2024-10-14 |
| | | |||
| * | Track errors when reading/substituting derivations | Christopher Baines | 2024-07-12 |
| | | |||
| * | Instrument the utility thread pool delays | Christopher Baines | 2024-07-12 |
| | | |||
| * | Increase the number of utility threads | Christopher Baines | 2024-07-12 |
| | | |||
| * | Split code for checking whether to store builds | Christopher Baines | 2024-07-09 |
| | | | | | | | To better separate out when a derivation is read, and when it's not. This should ensure that derivations aren't being read inside the transaction to submit builds. | ||
| * | Remove the workaround for with-store/non-blocking | Christopher Baines | 2024-07-04 |
| | | | | | As this is fixed in Guix now. | ||
| * | Revert "Add dbstat metrics" | Christopher Baines | 2024-07-04 |
| | | | | | | | It's not fast enough. This reverts commit 09daad84faea3577acf5f1627e4bfed699b10a5d. | ||
| * | Log the current state of the reader threads | Christopher Baines | 2024-07-02 |
| | | |||
| * | Track the procedures the worker threads are running | Christopher Baines | 2024-07-02 |
| | | |||
| * | Reduce the sleep in the worker threads | Christopher Baines | 2024-07-02 |
| | | |||
| * | Remove now redundant use of the writer thread channel | Christopher Baines | 2024-07-02 |
| | | | | | As there are no longer in memory tables. | ||
| * | Try not using the busy timeout | Christopher Baines | 2024-07-02 |
| | | | | | | I'm not quite sure what effect this will have, but it's probably better to do the waiting in Guile rather than SQLite. | ||
| * | Add dbstat metrics | Christopher Baines | 2024-07-02 |
| | | |||
| * | Don't use a chunked response for the metrics | Christopher Baines | 2024-06-30 |
| | | |||
| * | Add option to skip updating derived priorities when submitting builds | Christopher Baines | 2024-06-25 |
| | | | | | | As I think this can be skipped if you're submitting a batch of interdependent builds with the same priority. | ||
| * | Fix build prioritisation | Christopher Baines | 2024-06-25 |
| | | |||
| * | Add a utility thread pool | Christopher Baines | 2024-06-25 |
| | | | | | And use this to read derivations. | ||
| * | Remove support for chunked requests | Christopher Baines | 2024-06-23 |
| | | | | | | This was a hack to work around reading the entire request/response body in to memory, and is no longer needed. | ||
| * | Add missing loop when fetching builds to allocate | Christopher Baines | 2024-06-23 |
| | | |||
| * | Don't abuse the reader threads to read derivations | Christopher Baines | 2024-06-23 |
| | | | | | To avoid causing any issues with them. | ||
| * | Return copies of the list from the allocation plan fiber | Christopher Baines | 2024-06-22 |
| | | | | | To avoid the authoritative data structures from being modified. | ||
| * | Remove the ATTACH DATABASE line | Christopher Baines | 2024-06-22 |
| | | | | | As it's no longer used. | ||
| * | Add some logging in to build-coordinator-fetch-build-to-allocate | Christopher Baines | 2024-06-22 |
| | | |||
| * | Add some logging to the hook work queues | Christopher Baines | 2024-06-19 |
| | | |||
| * | Move the allocation plan in to memory | Christopher Baines | 2024-06-19 |
| | | | | | | Rather than using an in memory SQLite table, since I think this might be causing issues when opening new SQLite reader threads. | ||
| * | Reduce the nar size threshold | Christopher Baines | 2024-06-15 |
| | | | | | Unreasoned, but this'll compress more nars. | ||
| * | Change with-port-timeouts to use seconds for the timeout | Christopher Baines | 2024-06-10 |
| | | |||
| * | Validate derivation names for builds | Christopher Baines | 2024-06-09 |
| | | | | | As I've managed to submit derivation names with \r in. | ||
| * | Don't ignore missing_inputs setup failures when allocating | Christopher Baines | 2024-06-07 |
| | | | | | | | | | | As if there is an issue substituting the input, this can cause the build to fail to start over and over again, potentially retrying downloading the inputs over and over again. This will mean that the setup failures will need clearing manually before the build can be tried again, but this is probably a good thing. | ||
| * | Avoid using select and use poll instead | Christopher Baines | 2024-06-07 |
| | | | | | | I don't trust select, as I think I'm seeing it make blocking read calls. I don't know if that's coming from here, but better to just avoid it. | ||
| * | Support controlling adding timestamps to the log output | Christopher Baines | 2024-06-06 |
| | | | | | So this can be turned off when the shepherd is adding the timestamps. | ||
| * | Support customising the default-build-missing-inputs-hook behaviour | Christopher Baines | 2024-06-06 |
| | | |||
| * | Retry more when substituting derivations | Christopher Baines | 2024-06-06 |
| | | | | | And don't use the no substitutes fallback. | ||
| * | Always use a thread to read derivations | Christopher Baines | 2024-06-06 |
| | | |||
| * | Close the uncompressed output port explicitly | Christopher Baines | 2024-06-05 |
| | | | | | | As this ensures the data is written. This resolves some issues with bytes missing from the end of nars. | ||
| * | Make imports from (guix build utils) explicit | Christopher Baines | 2024-06-05 |
| | | |||
| * | Omit none as a compression in narinfos | Christopher Baines | 2024-06-05 |
| | | | | | To match the guix publish behaviour. | ||
| * | Enable recompressing nars in the publish hook | Christopher Baines | 2024-05-19 |
| | | | | | | | | | | | The implementation is intended to be generic, but the new default behaviour either keeps the lzip compressed nar, or if the compressed nar isn't much smaller than the compressed one, decompresses it. This also changes the API for the post-publish-hook and combined-post-publish-hook, since they now get called with a list of narinfos, rather than a list of narinfos and files. | ||
| * | Use threads for reading derivations | Christopher Baines | 2024-05-15 |
| | | | | | As it is quite slow on bayfront at least. | ||
| * | Work around the broken with-store/non-blocking | Christopher Baines | 2024-05-15 |
| | | |||
| * | Use just one cooperative thread | Christopher Baines | 2024-05-15 |
| | | | | | To try and simplify things. | ||
| * | Order messages to the writer thread channel | Christopher Baines | 2024-05-15 |
| | | | | | So that they're processed in a first come first served manor. | ||
| * | Handled fetching details of canceled builds | Christopher Baines | 2024-05-15 |
| | | |||
| * | Export port timeout error predicates | Christopher Baines | 2024-05-15 |
| | | |||
| * | Use with-store/non-blocking from guix | Christopher Baines | 2024-05-14 |
| | | | | | | As it opens a non-blocking connection, rather than doing a blocking handshake then making the socket non-blocking. | ||
| * | Fix the no-fibers-wait in with-fibers-port-timeouts | Christopher Baines | 2024-05-14 |
| | | |||
| * | Make the worker thread timeout a parameter | Christopher Baines | 2024-05-12 |
| | | |||
| * | Avoid starting new builds when low on disk space | Christopher Baines | 2024-05-12 |
| | | |||
| * | Increase the timeout for substituting derivations | Christopher Baines | 2024-05-10 |
| | | |||