| 23 Jan 2019 |
 _hc | actaully, never mind | 11:22:16 |
| _hc | we don have a way to pass that to the buildserver guest, so it doesnt matter much | 11:22:30 |
| _hc | but the RAM stuff is right | 11:22:39 |
 Bubu | _hc: the ram is set by the boostrap project: https://gitlab.com/fdroid/fdroid-bootstrap-buildserver/blob/master/Vagrantfile#L17 | 11:24:31 |
| Bubu | and uniq already upped that from 3 to 4 GB a while ago | 11:25:31 |
| _hc | Bubu: then your buldserver guest can have max RAM of the host minus 2GB | 11:33:27 |
| Bubu | which is set here and also to 4 GB | 11:37:32 |
| Bubu | which doesn;t sound like a very good idea 😅 | 11:37:47 |
| Bubu | maybe I should just run the whole thing from a ramdisk to get any resemblence of io performance 🤔 | 12:30:00 |
| Bubu | but I think there's not quite enough ram left for this | 12:31:38 |
| Bubu | _hc: there's a android-rebuilds and a gitlab runner vm, both reserving 32 GB of ram. are they in use right now? Could I stop them for today? | 13:09:29 |
| Bubu | btw, makebuildserver now continues to run. it was indeed the bad ram configuration. | 13:13:25 |
|  andybalaam joined the room. | 13:28:03 |
| _hc | Bubu: the defaults for that Vagrantfile and makebuildserver should align in a useful way, please submit a merge request if they don't | 14:40:05 |
| _hc | Bubu: gcc137 has about 55GB of RAM free | 14:40:40 |
| _hc | i.e. used for disk caching | 14:40:49 |
| _hc | `for f in `vagrant global-status | grep running | awk '{print $5}'`; do grep -i mem $f/Vagrantfile; done` | 14:42:54 |
| Bubu | In reply to @eighthave:matrix.org
Bubu: gcc137 has about 55GB of RAM free I know. which is not enough for running the whole bootstrap process in a ramdisk | 14:47:49 |
| Bubu | 110 gb just might be | 14:47:56 |
| Bubu | so if the two 32 gb boxes are currently not in use, I could try to use that ram | 14:49:23 |
| _hc | run it overnight | 14:58:26 |
| _hc | it takes about 5 hours | 14:58:27 |
| _hc | I've run it like 5ish times | 14:58:28 |
| _hc | I have zero time to do anything extra with managing VMs | 14:58:34 |
| _hc | that box has an SSD, so a RAM disk probably won't speed it up much | 14:59:29 |
| Bubu | I'm trying to fix makebuildserver. So it not taking 5 hours would make that process 100 times less annoying. | 15:27:54 |
| Bubu | I'll try to find out of we can have 1 or 2 more of those servers. | 15:28:17 |
| Bubu | And I wasn't really asking for you to manage anything. But you are probably the one who knows what these machines are currently used for (or if they aren't.) | 15:30:07 |
| _hc | the ones that are not in use are shutdown | 16:13:36 |
| _hc | the eschalot one isn't so important but it uses 512MB of ram | 16:14:21 |
