qiujunting | hello | 01:19 |
---|---|---|
qiujunting | The current cloud-init documentation is for users, is there any documentation for developers? | 01:19 |
qiujunting | I want to know how cloud-init is started? And how do you get the metadata? | 01:22 |
qiujunting | Hello | 08:04 |
qiujunting | I want to get a document about cloud-init for developer | 08:05 |
qiujunting | who can help me? | 08:06 |
qiujunting | where is a document about cloud-init for developer? | 11:34 |
qiujunting | Documents about cloud-init source code analysis? | 11:35 |
AnhVoMSFT | rharper: continuing on the mount -a" | 15:55 |
AnhVoMSFT | rharper: continuing on the "mount -a" discussion last week. if the user adds nfs mounts to the "mounts" (#cloud-config), given the code today, it won't work anyway on RHEL/CentOS/SUSE due to the dependency of rpcd. If we use mount -a -O no_netdev, we can document that if the user still wants to mount NFS devices, they should not annotate the mounts with netdev in #cloud-config | 15:57 |
rharper | AnhVoMSFT: sure; that's a bug ... where I left off was a suggestion to address the bug; there are two paths; 1) upstream RHEL could look to relax the restriction as other OSes have done w.r.t rpcd 2) for legacy images, cloud-init could look to include the drop-in script which does a mount -a -O _netdev if _netdev entries in exist in fstab | 16:00 |
AnhVoMSFT | rharper: for #2, my worry is that it's going to be hard to get it right. How do we know when rpcd will have run so that we can safely run mount -a -O _netdev? | 16:01 |
rharper | AnhVoMSFT: cloud-config.service already runs after networking is up After=network-online.target; so using a ExecStartPre= script means right before invoking cloud-init modules --mode=config; it would run the script to check for _netdev entries in /etc/fstab and do a mount -a -O _netdev; | 16:03 |
rharper | I suspect to be completely save, we'd want to also include an After=rpcd.service on the RHEL distros | 16:03 |
AnhVoMSFT | rharper: another option is to add a new configuration option to mounts to allow user/image builders to indicate whether they want cloud-init to mount netdev devices during "mounts" | 16:10 |
rharper | well, we already support nfs mounts, so users then need to know that if they use RHEL and nfs mounts, they have to pass some user-data config to tell cloud-init not to attempt to do nfs mounts (which may also be provided in the user-data config?) | 16:20 |
hggdh | Red Hat, CentOS, and Oracle Linux are affected | 18:34 |
rharper | hggdh: yeah | 18:56 |
powersj | Odd_Bloke, "we aren't aware of any distros backporting cloud-init to releases which use Python 3.4," the reason we decided on 3.4 was SuSE | 19:26 |
smoser1 | can i get someone to ack this: https://github.com/canonical/cloud-init/pull/879 | 20:56 |
=== smoser1 is now known as smoser | ||
Odd_Bloke | I don't entirely trust this smoser1 imposter, but I've +1'd anyway. ;) | 21:01 |
smoser | Odd_Bloke: so is that fine to land then? i just hit squash and merge ? | 23:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!