=== rangerpbzzzz is now known as rangerpb | ||
smoser | rharper, fyi, shadowusers really sucks. | 15:59 |
---|---|---|
smoser | s/shadowusers/extrausers/ | 15:59 |
smoser | http://pad.lv/1679765 | 15:59 |
smoser | bug 1679777 | 16:00 |
smoser | bug 1679765 | 16:00 |
smoser | i thought i had amup here. | 16:00 |
rharper | the first bug is a coreutils issue; I believe the --extrausers was introduced in some of those tools; I guess not all of them | 16:15 |
rharper | the second issue is related to cloud-init's concept of 'system-user'; that does need to be addressed; | 16:15 |
smoser | generally agree. | 16:16 |
smoser | the first bug is just a half-done implementation | 16:16 |
smoser | but so is the second. | 16:16 |
smoser | there are well defined ways that you can iterate users or get information about users. | 16:17 |
smoser | python uses such ways | 16:17 |
smoser | and they dont work with 'extrausers' | 16:17 |
rharper | it may end up being the same bug 'incomplete extrausers implementation' | 16:18 |
rharper | I suspect it's broken without cloud-init; ie, if you snap create-user with an assertion, the same issues remain w.r.t user manipulation/configuration/enumeration | 16:18 |
rharper | we should likely sync with foundations | 16:19 |
smoser | well, yes. cloud-init *does* have the ability to know that for users it created. | 16:21 |
smoser | but "what is the home directory for bob". should not require "is bob an extra user?" | 16:21 |
rharper | correct, which to me is a bug in the getent failure; I think if --extrausers is available in the tools, the the default search should try both paths (/etc/passwd , /var/lib/extrausers/passwd) | 16:31 |
rharper | etc | 16:31 |
harlowja | smoser hey, for the reporting stuff in cloud-init, what's a common usage of it? | 17:40 |
smoser | to do the bidding of harlowja | 17:41 |
harlowja | some folks at godaddy are trying to gather more insight into how long modules in cloud-init run for and ... | 17:41 |
smoser | but it was added for maas | 17:41 |
harlowja | and this reporting stuff looks designed for doing that (sort of) | 17:41 |
smoser | maas wanted status | 17:41 |
harlowja | ya, makes sense | 17:41 |
harlowja | sounds similar | 17:42 |
smoser | but yeah. the reporting stuff is doing that too | 17:42 |
harlowja | cool beans | 17:42 |
smoser | and rharper has stuff that reads a journal | 17:42 |
harlowja | ? | 17:42 |
harlowja | out loud? | 17:42 |
smoser | and timestpams things | 17:42 |
harlowja | cool cool | 17:42 |
smoser | it reads harlowja's journal | 17:42 |
harlowja | my diaryyyyy | 17:42 |
harlowja | noooo | 17:42 |
smoser | "Dear diary, today I met a new friend" | 17:42 |
harlowja | his name was bob | 17:42 |
smoser | "I think i love cloud-init" | 17:42 |
harlowja | lol | 17:42 |
harlowja | the things i write about smoser ... | 17:43 |
harlowja | don't read those | 17:43 |
harlowja | lol | 17:43 |
smoser | rharper, where do we see cloud-init annotate ? | 17:43 |
rharper | https://lists.ubuntu.com/archives/ubuntu-server/2016-October/007419.html | 17:43 |
rharper | https://code.launchpad.net/~raharper/+git/cloudinit-analyze | 17:43 |
harlowja | hmmm | 17:44 |
harlowja | interesting | 17:45 |
harlowja | cool, thx | 17:45 |
harlowja | forwarding some of htis | 17:45 |
rharper | also, it'd be highly useful if we wrapped each subp with a reporting event | 17:45 |
smoser | not the journal entries i hope | 17:45 |
rharper | that'd give us per exec data | 17:45 |
rharper | smoser: lol | 17:45 |
harlowja | whats this 'ubuntu-server' ml | 17:45 |
harlowja | :-P | 17:45 |
rharper | I also sent it to the cloud-init ML | 17:48 |
rharper | which is hosted on launchpad | 17:48 |
harlowja | oh | 17:48 |
* harlowja shuts up now | 17:48 | |
harlowja | lol | 17:48 |
erick3k | hi | 18:07 |
erick3k | can someone help me debug a timeout | 18:07 |
erick3k | please | 18:07 |
erick3k | https://paste.fedoraproject.org/paste/VifD~R6n8t78gj-m-Iv4CV5M1UNdIGYhyRLivL9gydE= | 18:07 |
smoser | erick3k, at very least you'll need to paste /var/log/cloud-init.log and also journalctl output | 18:09 |
erick3k | ty | 18:09 |
erick3k | https://paste.fedoraproject.org/paste/7rV6y2UBLmcJ77KPWP4PjF5M1UNdIGYhyRLivL9gydE=/raw | 18:12 |
erick3k | that is /var/log/cloud-init.log (although there is nothing other than a warning that config drive was unplugged) | 18:12 |
erick3k | and journalctl output says invalid argument | 18:13 |
=== rangerpb is now known as rangerpbzzzz | ||
erick3k | can someone help me when get a change, https://paste.fedoraproject.org/paste/6-WoDBw8~Pw9QtOd2KV~lF5M1UNdIGYhyRLivL9gydE= | 22:26 |
erick3k | getting big timeouts | 22:26 |
nacc | erick3k: which line? | 22:32 |
erick3k | nacc i dont really care about the warnings if they are not causing the timeout nor i care it can't print to the console but i am guessing this is the big timeout | 22:33 |
erick3k | Cloud-init v. 0.7.5 finished at Thu, 06 Apr 2017 22:16:28 +0000. Datasource DataSourceConfigDrive [local,ver=2][source=/dev/sr1]. Up 311.33 seconds | 22:33 |
erick3k | that is 5 mins | 22:33 |
erick3k | i guess? | 22:33 |
nacc | are these from multiple installs? why is cloud-init 'finished' more than once | 22:34 |
nacc | also it's a bit odd that the timezone appears to jump in the logs? | 22:35 |
nacc | 2017-04-06 18:16:28,208 | 22:35 |
nacc | Thu, 06 Apr 2017 22:16:28 +0000. | 22:35 |
nacc | lines 120 and 121 | 22:35 |
erick3k | ummm maybe i didn't clear the log | 22:35 |
erick3k | yes its weird and i can't find why those huge timeouts | 22:36 |
nacc | starts at Thu, 06 Apr 2017 22:11:22 | 22:36 |
erick3k | like the vm takes after is up, like 20 mins before it reboots (runcmd) | 22:36 |
nacc | ends at Thu, 06 Apr 2017 22:16:28 +0000 | 22:36 |
nacc | erick3k: which log is this? | 22:37 |
nacc | 2017-04-06 18:11:26,935 - cloud-init[WARNING]: Stdout, stderr changing to (>> /var/log/cloud-init-output.log, >> /var/log/cloud-init-output.log) | 22:37 |
erick3k | yes | 22:37 |
nacc | Cloud-init v. 0.7.5 running 'init' at Thu, 06 Apr 2017 22:16:21 +0000. Up 305.12 seconds. | 22:37 |
nacc | there is the 5 minute gap | 22:37 |
erick3k | changes cloud-init.log to cloud-init-output.log | 22:37 |
nacc | (i think) | 22:37 |
erick3k | not sure why, am using centos 7 cloud image | 22:37 |
erick3k | and why that gap? how can i tune that timeout thingy | 22:38 |
erick3k | any ideas nacc? | 22:39 |
nacc | erick3k: so that output is from cloud-init.log, erick3k ? | 22:40 |
erick3k | there is nothing on cloud-init.log | 22:40 |
nacc | erick3k: i'm not sure why you think it's a timeout? | 22:40 |
erick3k | well what could it be doing for 15 minutes? | 22:41 |
nacc | erick3k: there is a delay, but that doesn't necessarily equate to a timeout, unless you have some other data? | 22:41 |
nacc | 5 minutes, i thought? | 22:41 |
erick3k | na | 22:41 |
erick3k | once the machine is up and ready for login | 22:41 |
erick3k | last command in runcmd: -reboot is run like 15 - 20 minutes later | 22:41 |
nacc | erick3k: but i can only go off what's in the logs | 22:42 |
nacc | erick3k: you might need smoser or rharper for this | 22:43 |
erick3k | yes he tried but busy | 22:43 |
erick3k | thats all there is on logs, cloud-init.log is changed to cloud-init-output.log | 22:43 |
erick3k | so is same file | 22:43 |
nacc | erick3k: https://lists.ubuntu.com/archives/ubuntu-server/2016-October/007419.html | 22:43 |
nacc | erick3k: i wonder if that profiler would help | 22:44 |
erick3k | he saying 13% improvement | 22:45 |
nacc | erick3k: by making changes based upon profiling | 22:45 |
nacc | erick3k: my point was the profiler might tell you where all your time is being spent, if it's cloud-init | 22:45 |
erick3k | i have the machines in raid 0 with two NVME intel latest pci ssd, so you can imagine how fast things are | 22:45 |
erick3k | it boots in like 3 seconds so 15 minutes is just beyond ridiculous amount of time to have an instance ready | 22:46 |
erick3k | you think it could be a timeout on the os? | 22:47 |
erick3k | *OS | 22:47 |
nacc | erick3k: i really don't know -- was hoping it'd be something obvious, but i don't know enough about cloud-init or cento | 22:47 |
nacc | *centos | 22:47 |
erick3k | kind of same happens in ubuntu 14 | 22:48 |
erick3k | takes a long time | 22:48 |
erick3k | now that you tell me might be OS timeout for network gonna check that first | 22:48 |
erick3k | is there a way to have the network not even look for dhcp while turning on until cloudinit injects the network? | 22:49 |
nacc | erick3k: i assumethat would be centos specific | 22:49 |
erick3k | all | 22:50 |
erick3k | ubuntu / centos | 22:50 |
erick3k | i seal the vm with this in interfaces | 22:50 |
erick3k | DEVICE=eth0 | 22:51 |
erick3k | TYPE=Ethernet | 22:51 |
erick3k | BOOTPROTO=dhcp | 22:51 |
erick3k | ONBOOT=yes | 22:51 |
erick3k | what if i delete that? | 22:51 |
erick3k | can cloud-init still inject? | 22:51 |
erick3k | YES SR | 23:13 |
erick3k | THAT WAS THE TIMEOUT | 23:13 |
erick3k | FINALLY | 23:13 |
erick3k | hehe | 23:13 |
erick3k | now instance ready in 6 seconds | 23:14 |
erick3k | thats how its done :) | 23:14 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!