=== Lord-Kam_ is now known as Lord-Kamina | ||
RikMills | sil2100 seb128: hi, can I draw your attention to this: | 09:06 |
---|---|---|
RikMills | [09:51] <sitter> btw maybe someone can find someone to poke about https://bugs.launchpad.net/ubuntu/+source/libpwquality/+bug/1834480 it's gonna be impacting plasma 5.17 when I land https://phabricator.kde.org/D22122 | 09:07 |
ubottu | Launchpad bug 1834480 in langpack-o-matic "translations in not so ideal language-pack" [Undecided,New] | 09:07 |
=== alan_g is now known as alan_g_ | ||
ahasenack | tjaalton: hi, around? Could you please take a quick look at this paste: https://pastebin.ubuntu.com/p/ZzW8BG2fpm/ line 16 | 13:07 |
ahasenack | it seems to imply there is an invisible default /etc/sssd/sssd.conf file | 13:07 |
ahasenack | I'm checking in the debian build now to compare | 13:07 |
ahasenack | same in debian | 13:10 |
tjaalton | ahasenack: and this is with the latest 2.2.0? | 13:18 |
ahasenack | yes | 13:18 |
ahasenack | the install doesn't fail | 13:18 |
ahasenack | tests pass | 13:18 |
ahasenack | but the socket services try to start, and fail like that | 13:19 |
ahasenack | I'm wondering if they could use a ConditionFileNotEmpty | 13:19 |
ahasenack | I guess fedora/rh don't see this because they don't start services by default? | 13:19 |
ahasenack | I'm ready to send an email to sssd-users@ | 13:19 |
tjaalton | well no-one has complained about it on debian | 13:20 |
tjaalton | I don't know why it thinks sssd.conf is there | 13:20 |
ahasenack | it also happens there, this is the postinst output: | 13:20 |
ahasenack | tjaalton: https://pastebin.ubuntu.com/p/5cSFT4hmnJ/ | 13:21 |
tjaalton | so it needs sssd.conf? | 13:22 |
tjaalton | I haven't used it in a while.. | 13:23 |
ahasenack | a few other services start | 13:23 |
ahasenack | these 3 | 13:23 |
ahasenack | 1871 ? Ss 0:00 /usr/sbin/sssd -i --logger=files | 13:23 |
ahasenack | 1872 ? S 0:00 \_ /usr/libexec/sssd/sssd_be --domain implicit_files --uid 0 --gid 0 --logger=files | 13:23 |
ahasenack | 1873 ? S 0:00 \_ /usr/libexec/sssd/sssd_nss --uid 0 --gid 0 --logger=files | 13:23 |
ahasenack | probably noops | 13:23 |
tjaalton | there's your nss | 13:24 |
ahasenack | but sssd-nss.service stays it's disabled, is this the same? | 13:25 |
ahasenack | execstart for sssd-nss.service has "ExecStart=/usr/libexec/sssd/sssd_nss ${DEBUG_LOGGER} --socket-activated" | 13:25 |
tjaalton | did you have sssd running before the upgrade? | 13:26 |
ahasenack | nope, fresh install in debian-sid | 13:26 |
ahasenack | (same case in ubuntu) | 13:27 |
ahasenack | sssd spawned sssd_nss, but there is also a systemd service file for sssd-nss | 13:27 |
ahasenack | is this a case where one would choose one or the other? Either start it from the get-go, or use socket activations? | 13:28 |
ahasenack | I'm thinking the default implicit config (without sssd.conf) assumes it should start, and we are at the same time trying to use socket activation, but the socket activation is a bit incompatible with this implicit ssd.conf | 13:29 |
tjaalton | I need to test on a vm | 13:31 |
tjaalton | sounds like it should use different defaults when starting without a conffile | 13:34 |
tjaalton | or not start at all | 13:37 |
ahasenack | we could add a ConditionFileNotEmpty, or whatever it's called | 13:37 |
ahasenack | to the socket systemd services | 13:37 |
ahasenack | but it sounds like two categories of services: the normal ones, where you just start it even if it's not used,and the socket activated ones | 13:38 |
ahasenack | starting both seems wrong | 13:38 |
tjaalton | socket activated ones shouldn't be "started" | 13:39 |
tjaalton | so the postinst output looks weird to me | 13:39 |
ahasenack | if we do nothing against it, dh_installsystemd will enable it on first install | 13:42 |
tjaalton | meh, when is samba-common-bin postinst going to be fixed to not fail when smb.conf isn't around | 13:43 |
ahasenack | hm, "The services' list is optional on platforms where systemd is supported, as they will either be socket or D-Bus activated when needed." | 13:51 |
tjaalton | I was wrong, the socket listener is actually started and is shown by systemctl | 13:53 |
tjaalton | anyway, with the conffile from examples it'll work properly, so the daemon defaults are just wrong | 13:53 |
ahasenack | you mean defaults without a conffile | 13:55 |
tjaalton | yep | 13:56 |
tjaalton | upstream says it shouldn't start with sssd_nss | 13:58 |
ahasenack | it's the same about all others | 13:58 |
ahasenack | Aug 01 12:49:07 eoan-sssd2 sssd_check_socket_activated_responders[3014]: The sudo responder has been configured to be socket-activated but it's still mentioned in the services' line in /etc/sssd/sssd.conf. | 13:59 |
ahasenack | sudo, for example | 13:59 |
tjaalton | yes, but for the same reason | 13:59 |
tjaalton | with a dummy conffile it's fine | 13:59 |
ahasenack | so "services" suddenly grew a long list of defaults if there is no sssd.conf | 14:00 |
tjaalton | something like that | 14:00 |
ahasenack | I emailed sssd-users@ | 14:00 |
tjaalton | I'm talking with lslebodn | 14:00 |
ahasenack | who is that? | 14:00 |
ahasenack | and which channel, assuming it's irc? | 14:01 |
tjaalton | #sssd | 14:01 |
tjaalton | lukas slebodnik | 14:01 |
tjaalton | from redhat | 14:01 |
tomreyn | is 18.04.3 likely to happen today, i.e. are there known reasons making it unlikely right now (i'm aware such can come up later)? | 14:11 |
tjaalton | tomreyn: no, delayed for one week | 14:12 |
tomreyn | thanks tjaalton , did i miss an e-mail on this? or will there likely be one? | 14:12 |
tjaalton | uh, there isn't one? | 14:13 |
tomreyn | not on -announce or -release | 14:13 |
tjaalton | ok then | 14:14 |
tjaalton | well that's what I heard somewhere | 14:14 |
tjaalton | maybe I'm wrong | 14:14 |
tomreyn | i'm looking at https://lists.ubuntu.com/archives/ubuntu-announce/ | 14:14 |
tomreyn | is this the right place? | 14:14 |
analogical | tjaalton, how do you know that 18.04.3 is delayed for a week?? | 14:14 |
tjaalton | I don't | 14:15 |
tjaalton | it was a rumor it seems | 14:15 |
tomreyn | sorry to hear this ;) | 14:16 |
tjaalton | ahasenack: so, the verdict is that socket activation just doesn't work without a conffile right now. but it doesn't actually break anything, once sssd has been configured and restarted it's all fine | 14:27 |
tjaalton | trying to find where the defaults are set | 14:28 |
tjaalton | should be a simple patch | 14:28 |
ahasenack | should we add a ConditionFileNotEmpty? | 14:28 |
ahasenack | for sssd.conf | 14:28 |
tjaalton | service you mean? | 14:29 |
tjaalton | sssd.service | 14:30 |
ahasenack | I mean to the systemd socket files | 14:30 |
ahasenack | so they won't start without a config file | 14:30 |
ahasenack | sssd starts fine, albeit useless | 14:31 |
tjaalton | I won't do that in debian ;) | 14:31 |
ahasenack | no prior art? | 14:31 |
tjaalton | looks like src/confdb/confdb_setup.c needs fixing | 14:31 |
tjaalton | it's a non-issue mostly | 14:31 |
tjaalton | install is noisy, that's all | 14:32 |
ahasenack | and if someone checks the status output, as suggested in that output | 14:32 |
ahasenack | but nothing broke, as evidenced by the tests even | 14:32 |
tjaalton | I'll patch the fallback setup instead | 14:34 |
ahasenack | char fallback_cfg[] = ... <-- that bit | 14:35 |
gaughen | Apologies all, the point release is delayed. | 14:35 |
ahasenack | although it's not listing sudo and the others | 14:35 |
gaughen | it will not happen today, there were issues found in kernel testing. | 14:35 |
tjaalton | to match the patched examples/sssd.conf | 14:35 |
tjaalton | ahasenack: I think socket activation breaks if services lists any of them | 14:35 |
ahasenack | yes, it's one or the other | 14:36 |
ahasenack | but this implicit config, when there is no config file, isn't listing sudo or the others, and they are still complaining that they were listed | 14:36 |
tjaalton | the output just assumes there's a conffile with services= | 14:36 |
tjaalton | tomreyn: ^ the release is in fact delayed | 14:38 |
tomreyn | tjaalton, gaughen: thanks for the notice - can you say until when it'll be postponed, yet, should users check ubuntu-announce for a notice on this? | 14:40 |
gaughen | tomreyn, tjaalton last I knew the plan was for next Thursday, but infinity would have the latest info has he's running the release | 14:41 |
gaughen | tomreyn, I do not think that's an unreasonable expectation, apologies for dropping this on y'all at the last minute | 14:41 |
tomreyn | personally i'm always happy with delays if it helps stabilizing a release. | 14:43 |
tjaalton | ahasenack: patching it didn't help for some reason | 15:28 |
ahasenack | tjaalton: you just dropped the services line? | 15:33 |
ahasenack | in that fallback config | 15:33 |
tjaalton | yes | 15:33 |
tjaalton | then I created a conffile which matches the fallback, and then it works.. so looks like the checker is broken | 15:34 |
tomreyn | so, since it's EOB in UK, should we expect an e-mail or other announcement about the 18.04.3 delay? | 16:02 |
vorlon | there will be an email announcement; the release manager is not on UK time | 16:03 |
tomreyn | great :) | 16:04 |
tjaalton | ahasenack: check_socket_activated_responder() checks for the conffile, and if it doesn't exist, bails out | 16:07 |
tjaalton | ahasenack: I'm currently bisecting a kernel for a regression found in 5.0, so if you have time to poke sssd further then feel free to ;) | 16:08 |
LocutusOfBorg | hello, broadcast question: (I'm talking about flang): how can a package that *never* installs libomp5 in the build process runtime depend on it? | 16:38 |
LocutusOfBorg | Depends: libc6 (>= 2.29), libgcc1 (>= 1:3.0), libomp5 (>= 0.20130412) | 16:38 |
ahasenack | tjaalton: thanks for the troubleshooting done so far | 16:47 |
=== karimsye_ is now known as karimsye | ||
=== JamieBennett_ is now known as JamieBennett | ||
=== tedg_ is now known as tedg | ||
=== jamespage_ is now known as jamespage | ||
=== kenvandine_ is now known as kenvandine | ||
=== zyga_ is now known as zyga | ||
=== odc_ is now known as odc | ||
=== joedborg_ is now known as joedborg |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!