| *** toscalix has quit IRC | 00:10 | |
| *** mohan43u has quit IRC | 03:33 | |
| *** tristan has joined #buildstream | 04:38 | |
| *** mohan43u has joined #buildstream | 05:48 | |
| *** tristan has quit IRC | 06:06 | |
| *** seanborg__ has joined #buildstream | 07:17 | |
| *** benschubert has joined #buildstream | 07:36 | |
| *** jude has joined #buildstream | 07:44 | |
| *** toscalix has joined #buildstream | 08:02 | |
| *** tpollard has joined #buildstream | 08:11 | |
| *** santi has joined #buildstream | 08:24 | |
| *** phildawson has joined #buildstream | 08:24 | |
| *** lachlan has joined #buildstream | 09:06 | |
| *** lachlan has quit IRC | 09:13 | |
| *** finnb has joined #buildstream | 09:14 | |
| coldtom | has the 1.93.3.dev0 snapshot not been uploaded to PyPI? | 09:23 |
|---|---|---|
| *** lachlan has joined #buildstream | 09:27 | |
| coldtom | oh, 1.93.x for x > 0 hasn't been uploaded | 09:31 |
| *** lachlan has quit IRC | 10:02 | |
| valentind | I have noticed there is 1.93.3 now, but not for bst-plugins-experimental | 10:12 |
| valentind | Is there a plan to release it? | 10:13 |
| *** toscalix has quit IRC | 10:39 | |
| *** lachlan has joined #buildstream | 10:41 | |
| *** lachlan has quit IRC | 10:47 | |
| *** lachlan has joined #buildstream | 11:05 | |
| *** lachlan has quit IRC | 11:09 | |
| *** lachlan has joined #buildstream | 11:29 | |
| *** lachlan has quit IRC | 11:32 | |
| *** lachlan has joined #buildstream | 11:52 | |
| *** lachlan has quit IRC | 12:02 | |
| *** tristan has joined #buildstream | 12:02 | |
| *** pointswaves has joined #buildstream | 12:07 | |
| *** slaf has quit IRC | 12:22 | |
| *** lachlan has joined #buildstream | 13:17 | |
| *** lachlan has quit IRC | 13:21 | |
| *** lachlan has joined #buildstream | 13:35 | |
| *** lachlan has quit IRC | 13:38 | |
| *** lachlan has joined #buildstream | 13:52 | |
| *** lachlan has quit IRC | 13:59 | |
| *** slaf has joined #buildstream | 14:18 | |
| *** slaf has quit IRC | 14:19 | |
| *** devcurmudgeon has joined #buildstream | 14:23 | |
| *** slaf has joined #buildstream | 14:49 | |
| *** slaf has quit IRC | 14:51 | |
| bethw | +R | 14:52 |
| bethw | Whoops :) | 14:52 |
| *** slaf has joined #buildstream | 15:08 | |
| *** slaf has quit IRC | 15:09 | |
| *** lachlan has joined #buildstream | 15:13 | |
| *** lachlan has quit IRC | 15:20 | |
| *** doras[m] is now known as doras | 15:36 | |
| *** doras has quit IRC | 15:40 | |
| *** doras has joined #buildstream | 15:40 | |
| *** lachlan has joined #buildstream | 16:01 | |
| *** seanborg__ has quit IRC | 16:03 | |
| *** lachlan has quit IRC | 16:04 | |
| *** tristan has quit IRC | 16:07 | |
| *** tristan has joined #buildstream | 16:13 | |
| *** lachlan has joined #buildstream | 16:19 | |
| cphang | hey! I have a remote-execution question (or two). After an ExecuteResponse is returned does BuildStream expect the server to update the artifactCache? When retrieving cached results, will BuildStream probe the actionCache as well as the artifactCache? | 16:20 |
| *** lachlan has quit IRC | 16:25 | |
| *** mohan43u has quit IRC | 16:30 | |
| *** mohan43u has joined #buildstream | 16:33 | |
| *** mohan43u has quit IRC | 16:38 | |
| juergbi | cphang: 1) no, buildstream will push the artifact to the artifact server if configured. ideally, RE and artifact server use the same CAS server, though, to avoid having buildstream to upload the bulk of the blobs | 16:39 |
| juergbi | 2) artifact servers are probed as part of 'pull'. action cache is consulted if pull was not successful but only as part of the 'build' job, i.e., right before uploading the input blobs / submitting the Execute request | 16:41 |
| juergbi | action cache is optional. the RE server typically consults the action cache internally before execution. the benefit of configuring an action cache in BuildStream is that the input blobs are not uploaded if there is a hit in the action cache (at the cost of an extra round trip) | 16:42 |
| *** mohan43u has joined #buildstream | 16:42 | |
| *** jude has quit IRC | 16:47 | |
| cphang | juergbi makes sense tyvm. Scott has been doing work with buildbarn compatability with Buildstream and is encountering the following error in the ExecuteReponse https://gitlab.com/celduin/infrastructure/celduin-infra/-/issues/120#note_335096575 I'm wondering whether this is due to a missing artifactCache (it isn't present in the testing setup). Is | 16:50 |
| cphang | it possible BuildStream isn't uploading missing CAS blobs before the Execute request is issued? | 16:50 |
| juergbi | cphang: it should upload the complete input tree right before Execute by first checking what's missing with FindMissingBlobs and then using batch upload | 16:54 |
| cphang | indeed | 16:54 |
| juergbi | it does assume that in the short timespan between upload and Execute the blobs stay in CAS | 16:54 |
| cphang | indeed which is covered in the spec, We think in this case that in Scott's experimental conditions there is sufficient storage in the CAS. He's increased the capacity and still sees the same result. | 16:55 |
| juergbi | there could be a bug in the upload code somewhere, of course, maybe even in buildbox-casd which is used as local proxy | 16:55 |
| juergbi | if it's reproducible, it would probably make sense to check which digests are in the request/response of FindMissingBlobs and the uploads | 16:56 |
| cphang | juergbi on Monday I've asked Scott to further root cause to see whether my theory is correct/incorrect. We'll continue posting up against https://gitlab.com/celduin/infrastructure/celduin-infra/-/issues/120. If we can confirm it's a bug in buildstream we'll of course file an issue against that. | 16:56 |
| cphang | Scott can confirm it's reproducible when building against the same element | 16:57 |
| juergbi | enabling tracing for buildbox-casd and then checking cache/buildstream/logs/_casd could be useful | 16:57 |
| *** phildawson has quit IRC | 16:58 | |
| *** phildawson has joined #buildstream | 16:58 | |
| juergbi | buildstream log level 'debug' enables casd trace logs | 16:58 |
| *** tpollard has quit IRC | 16:59 | |
| juergbi | can be quite a lot of output but if it can be reproduced with a single element, it might be ok | 16:59 |
| *** phildawson has quit IRC | 17:00 | |
| *** phildawson has joined #buildstream | 17:00 | |
| cphang | mmm nice. | 17:01 |
| cphang | juergbi do you mind if I post up our discussion in https://gitlab.com/celduin/infrastructure/celduin-infra/-/issues/120 ? This has been incredibly helpful. | 17:01 |
| juergbi | go ahead | 17:02 |
| cphang | tyvm | 17:02 |
| *** phildawson has quit IRC | 17:03 | |
| *** santi has quit IRC | 17:21 | |
| *** mohan43u has quit IRC | 17:27 | |
| robjh | setting up a new gitlab runner to do buildstream things in docker. it looks like bubblewrap isnt working in the docker container. im using buildstream/buildstream:dev-extra as my image, and ive seen this get further on other runners. is there something i need to setup on the runner to make this work? a nesting option perhaps? | 17:41 |
| *** lachlan has joined #buildstream | 17:48 | |
| *** lachlan has quit IRC | 17:53 | |
| pointswaves | robjh, bouble wrap may need a privilaged container.. | 18:08 |
| *** lachlan has joined #buildstream | 18:08 | |
| pointswaves | i but i didnt think it did | 18:09 |
| robjh | im still assuming my host isnt setup right | 18:13 |
| *** lachlan has quit IRC | 18:15 | |
| jjardon | robjh: you need a privileged container, all buildstream runners are configured as privileged | 18:25 |
| robjh | giving that a try now | 18:34 |
| juergbi | yes, while bubblewrap doesn't need any special privileges on current mainline kernels, Docker seems to block it by default, unfortunately | 18:40 |
| jjardon | it would be nice to know what exact capabilities it needs, so we can open those and not everything | 18:44 |
| robjh | it looks like its got past the part it was failing at, thanks pointswaves, jjardon :) | 18:49 |
| jjardon | robjh: nice, good luck! | 18:49 |
| robjh | damn it, spoke too soon! | 18:53 |
| robjh | i edited the wrong config.toml file :p | 18:58 |
| *** toscalix has joined #buildstream | 19:10 | |
| *** phoenix has joined #buildstream | 19:21 | |
| valentind | Why do I get "This platform does not support local builds: buildbox-run not found" when it worked before? | 19:26 |
| *** tristan has quit IRC | 19:26 | |
| jjardon | valentind: maybe this recent change is related: new default is buildbox-run instead bwrap: https://gitlab.com/BuildStream/buildstream/-/merge_requests/1888 | 19:29 |
| valentind | With bwrap 0.3.3 | 19:29 |
| valentind | What is the correct way? Define BST_FORCE_SANDBOX? | 19:30 |
| valentind | Or is there a configuration file? | 19:30 |
| valentind | I set that for the moment. Maybe next week I will add buildbox in the docker image. | 19:33 |
| jjardon | valentind: I know almost nothing about this but seems BST_FORCE_SANDBOX=bwrap will be enough to still use bwrap, but seems bwarp backend is going to be removed soon so better try to fix the buildbox-run problem | 19:46 |
| *** benschubert has quit IRC | 19:55 | |
| juergbi | jjardon: no special capabilities are required on mainline kernel where user namespaces are enabled. docker might block it with seccomp, not sure. some distros also disable user namespaces by default, requiring a sysctl change | 20:39 |
| juergbi | valentind: BST_FORCE_SANDBOX is the only way but as jjardon writes, the old bwrap backend is going away, probably around end of this month. to be clear, the recommended buildbox-run implementation on Linux is buildbox-run-bubblewrap, i.e., the sandboxing approach doesn't change on the system level. you also need to install buildbox-fuse for safe operation as a single user | 20:42 |
| juergbi | also see https://buildstream.build/buildbox_install.html | 20:43 |
| *** pointswaves has quit IRC | 22:02 | |
| robjh | i would like to setup an artifact server for the runner i just set up. are there instructions for that? | 22:06 |
| jjardon | robjh: https://docs.buildstream.build/1.4.2/install_artifacts.html | 22:34 |
| jjardon | here for master: https://docs.buildstream.build/master/using_configuring_cache_server.html | 22:35 |
| robjh | jjardon, is 1.9 close enough to master for the master instructions to count? | 22:38 |
| jjardon | robjh: yes, I think so | 22:38 |
| valentind | Is buildbox-fuse needed? I do not have the impression it is. | 22:39 |
| robjh | jjardon, thanks | 22:41 |
| jjardon | valentind: this presentation https://docs.google.com/presentation/d/1tuqvro_Kpn1yLthqbsxImkcHaCKsGqsn2liuCwRsg7g/edit#slide=id.g5fefbbc030_0_0 says that is legacy | 22:50 |
| jjardon | (slide 7) | 22:50 |
| *** toscalix has quit IRC | 23:00 | |
| *** mohan43u has joined #buildstream | 23:53 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!