| === mpontillo_ is now known as mpontillo | ||
| === X-Istence is now known as x58 | ||
| === promethe1nfire is now known as prometheanfire | ||
| === cpaelzer_ is now known as cpaelzer | ||
| === rangerpbzzzz is now known as rangerpb | ||
| === roaksoax_ is now known as roaksoax | ||
| === daniel3 is now known as Odd_Bloke | ||
| smoser | harlowja_at_home, pitti (ubuntu systemd maintainer) is complaining about your Before=NetworkManager.service | 16:28 |
|---|---|---|
| pitti | well, not "complaining", but it should be a no-op | 16:29 |
| pitti | all network-y things like NM, networkd, ifupdown etc. already run After=network-pre.target -- that's the raison d'être for that target | 16:29 |
| pitti | so I was wondering if there is some race condition somewhere and that just hides a real bug | 16:30 |
| harlowja_at_home | nrezinor1, ^ | 16:37 |
| harlowja_at_home | x58, ^ | 16:37 |
| harlowja_at_home | apparently thats not how it works in RHEL | 16:37 |
| harlowja_at_home | lol | 16:37 |
| harlowja_at_home | those 2 folks are the experts here | 16:37 |
| harlowja_at_home | i'm just a poor small enginner man | 16:37 |
| harlowja_at_home | lol | 16:37 |
| harlowja_at_home | bbiab | 16:39 |
| harlowja | ok backs | 17:31 |
| nrezinor1 | i see my name was pinged lol , wat now | 17:45 |
| === nrezinor1 is now known as nrezinorn | ||
| harlowja | nrezinorn something about the network manager stuff | 18:03 |
| harlowja | nrezinorn https://irclogs.ubuntu.com/2016/09/26/%23cloud-init.html#t16:28 | 18:04 |
| nrezinorn | yeah i dont remember most of that haha | 18:13 |
| nrezinorn | that was so last week | 18:13 |
| harlowja | :-P | 18:13 |
| nrezinorn | the tl;dr is "MOAR TESTING" | 18:13 |
| nrezinorn | i dont have time to test things today lol stupid "real work" | 18:14 |
| x58 | smoser: RHEL requires Before=NetworkManager.service because otherwise NetworkManager starts at the same time as cloud-init-local, and takes over managing /etc/resolv.conf for instance. | 18:58 |
| x58 | smoser: I filed bugs for these issues, that describe what is going on. | 18:58 |
| x58 | RHEL doesn't have NetworkManger.service require network-pre or something that allows us to order cloud-init-local first. | 18:59 |
| x58 | pitti: Yeah, it would be nice if network-pre.target would actually be defined that way on RHEL/CentOS too... but alas it is not. | 18:59 |
| x58 | pitti: Also, networking.service is network.service on RHEL, so there are just difference between Ubuntu/RHEL. | 19:00 |
| x58 | Either split out the systemd per system, or template it someway, but those are things that need to be accounted for. | 19:00 |
| harlowja | agreed | 19:14 |
| harlowja | pitti if we need to split out thats fine | 19:14 |
| harlowja | if there is a RHEL bug somewhere, that'd be nice to know also | 19:14 |
| smoser | x58, so pitt was saying (i think) that upstream NetworkManager's service file does have network-pre.target | 19:18 |
| harlowja | in RHEL? | 19:18 |
| harlowja | or ubuntu | 19:18 |
| harlowja | ? | 19:18 |
| smoser | upstream | 19:18 |
| harlowja | upstream!!! | 19:18 |
| harlowja | :-P | 19:18 |
| harlowja | which stream | 19:18 |
| smoser | https://github.com/NetworkManager/NetworkManager/blob/master/data/NetworkManager.service.in | 19:20 |
| smoser | that one | 19:20 |
| harlowja | x58 ^ any idea, maybe RH just doesn't have that one yet | 19:23 |
| pitti | x58: right, forget the particular .service names; the point is that <stuff that fiddles with network interfaces> is After=network-pre.target, and <stuff that sets up firewall or config> runs before that | 19:53 |
| pitti | x58: and neither network-pre.target nor NetworkManager.service are an Ubuntu invention -- these both come from their upstream projects | 19:54 |
| pitti | x58, harlowja: so, it's plausible that RHEL's NetworkManager.service does not (yet) have After=network-pre.target, and then that workaround would actually DTRT; I just wanted to confirm that this is actually the case | 19:54 |
| pitti | because, if NM.service already does have After=network-pre.target, then that change is a no-op, and something is still not understood | 19:55 |
| apollo13 | are you guessing? cause I can look into centos7 if you want | 19:56 |
| pitti | I am just guessing, yes (Debian/Ubuntu guy here) | 19:59 |
| apollo13 | https://dpaste.de/UKgv/raw that is the current file on centos/rhel 7 | 19:59 |
| pitti | ah, so that indeed still misses it | 19:59 |
| apollo13 | anything else I should look at? | 20:00 |
| pitti | so that explains it (even though fixing the actual NM.service would be preferrable, so this ought to be a downstream patch) | 20:00 |
| pitti | but nevermind, I mostly just wanted to know if that was maybe just a misunderstanding | 20:01 |
| apollo13 | great | 20:01 |
| apollo13 | anyways, your After came in with nm 1.2, centos has 1.0.6 | 20:02 |
| apollo13 | https://bugzilla.gnome.org/show_bug.cgi?id=761001 was the original issue | 20:02 |
| apollo13 | although according to that bugreport it would be on 1-0 too? seems that redhat didn't cherrypick that (yet) | 20:03 |
| === rangerpb is now known as rangerpbzzzz | ||
| x58 | Yes it should probably be fixed by RHEL/CentOS, but that will probably take till version 193943834342112 ... | 22:32 |
| x58 | (being overly generous with that version number, knowing how slow RHEL moves) | 22:33 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!