*** radez has quit IRC | 01:03 | |
*** Sukhdev has quit IRC | 01:12 | |
*** Sukhdev has joined #openstack-dib | 01:22 | |
*** Sukhdev has quit IRC | 01:22 | |
*** Sukhdev has joined #openstack-dib | 02:36 | |
*** xinliang has quit IRC | 02:49 | |
*** xinliang has joined #openstack-dib | 03:01 | |
*** owalsh_ has joined #openstack-dib | 03:19 | |
*** owalsh has quit IRC | 03:23 | |
*** Sukhdev has quit IRC | 03:40 | |
*** jamielennox is now known as jamielennox|away | 03:46 | |
*** aparnav has joined #openstack-dib | 03:48 | |
*** jamielennox|away is now known as jamielennox | 03:53 | |
*** aparnav has quit IRC | 04:15 | |
*** chhavi has joined #openstack-dib | 04:34 | |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Switch to using "partition" objects https://review.openstack.org/465916 | 05:08 |
---|---|---|
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Move exception to it's own file (again) https://review.openstack.org/466129 | 05:08 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Move parts of Partition creation into object https://review.openstack.org/465838 | 05:08 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Split partition into it's own file https://review.openstack.org/465837 | 05:08 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Move to generic configuration parsing https://review.openstack.org/465417 | 05:08 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Switch debian to deb.debian.org https://review.openstack.org/466182 | 05:08 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove tree config parser from plugins https://review.openstack.org/466183 | 05:08 |
ianw | yolanda: ^ stack ready for review! | 05:09 |
ianw | i'd really like you to test with multiple partitions. as called out, we have lots of options for ordering. i'd like to get suggestions | 05:10 |
ianw | couple of mirror failures, but looks like CI is otherwise good with that stack | 05:52 |
ianw | which is good for the simple case at least | 05:52 |
andreas-f | ianw: just commented your partition objects patch. What do you think? | 05:54 |
ianw | andreas-f: so i don't think we need pre/post phases? | 05:55 |
ianw | the "partitioning" object just does the partitioning when it's create() gets called | 05:56 |
ianw | it just needs to know how to walk to the partition objects attached to it to get them in the right order? | 05:56 |
ianw | i'm fine with the "partition" objects being essentially no-ops ... their create() does nothing because they're created by their parent node | 05:56 |
ianw | so it's something like | 05:57 |
ianw | pre steps | 05:57 |
ianw | for node in sorted_partitions_list: partitions.append(node) | 05:57 |
ianw | do partitioning magic | 05:57 |
ianw | post steps | 05:57 |
andreas-f | Yes - the order is the problem: when we use your proposal we need an 'additional ordering' that has to be implemented in the partitioning (any maybe other) objects. | 05:58 |
andreas-f | When we use a more generic way (adding weights) this is solved once for all. | 05:58 |
ianw | right, so one simple way would be to just put a int() in each partition that is the order | 05:58 |
ianw | then, you walk the partitions, do a sort by that number, and you've got them in order | 05:58 |
andreas-f | Exactly. | 05:59 |
ianw | although, i'm not sure if it's not deterministic that when attaching a node, when you walk, do you get them back in the same order anyway? | 05:59 |
ianw | is it stable? | 05:59 |
ianw | it's not based on a hash? | 05:59 |
andreas-f | No - exactly this is the problem - which can be solved using weights (i.e. numbers). | 06:00 |
andreas-f | Instead of having the 'number' in the node - we put it into the edge. | 06:00 |
andreas-f | Then the algorithms using the graph can use this number to order them. | 06:01 |
ianw | how do we express that in the config? | 06:01 |
andreas-f | The config will only chance for partitions: the number has to be added. | 06:01 |
andreas-f | For all other (where the order is not important) we could use a default '0'. | 06:02 |
ianw | it seems the only change required here will be in line 163 of https://review.openstack.org/#/c/465916/4/diskimage_builder/block_device/level1/partitioning.py | 06:02 |
ianw | instead of using get_iter_outgoing() it can use something else, to get the weighted values in the right order? | 06:02 |
andreas-f | Not even this: the get_iter_outgoing() knows about weights (because they have to be implemented in the digraph level). | 06:04 |
andreas-f | and returns the outgoing nides in the correct way. | 06:04 |
andreas-f | The Interface for the create_edge() must change: an additional (defaulted) weight parameter must be added. | 06:05 |
ianw | andreas-f: ok, i'm pretty much out of time for today. i think we're pretty close to the same idea on this. if you have time to implement something like this i'd appreciate it as a follow-on to 465916 | 06:07 |
ianw | otherwise i'll get into it more next week my time | 06:07 |
ianw | we can always squash changes, etc, once we've reviewed them, but let's try to keep thing in very digestible chunks | 06:07 |
andreas-f | I also have to leave; I try to do something during the weekend. | 06:07 |
andreas-f | Yes - understand: only small changes ;-) | 06:08 |
ianw | andreas-f: ok, thanks. are you broadly ok with the idea of centralising the config like this? i feel like it results in much more consistent behaviour | 06:08 |
ianw | especially when we start adding on the lvm/pv/ etc etc stuff | 06:08 |
andreas-f | Yes - I'm find with this: it sounds really good. | 06:08 |
ianw | andreas-f: ok, great, well let's sync up later. cheers | 06:09 |
andreas-f | Bye. | 06:09 |
yolanda | hi ianw , looking at the conversation now | 06:22 |
yolanda | for sure i can test it | 06:22 |
*** andreas-f has quit IRC | 06:57 | |
*** adam_g has quit IRC | 07:00 | |
*** adam_g has joined #openstack-dib | 07:01 | |
*** adam_g has quit IRC | 07:12 | |
*** adam_g has joined #openstack-dib | 07:13 | |
*** adam_g has quit IRC | 07:21 | |
*** adam_g has joined #openstack-dib | 07:22 | |
ianw | yolanda: thanks ... as discussed we might need the edge weighting stuff for multiple partitions ... or maybe it just happens to work too. i dunno :) some of the earlier bits are probably ok as is, they're just code movement with no changes | 07:42 |
openstackgerrit | yolanda.robla proposed openstack/diskimage-builder master: Add dracut-regenerate elements https://review.openstack.org/447047 | 07:43 |
yolanda | so if the elements are on a list, they will likely preserve the order. The problem i was hitting before, that Andreas fixed, is that they were stored on a dictionary, so the order was random | 07:44 |
ianw | yeah, i'm not sure about the digraph implementation in the background | 07:45 |
ianw | that was why i suggested it might "just work" anyway if the list ordering is preserved, even if that's actually unintentional | 07:45 |
*** makowals_ has quit IRC | 08:34 | |
*** xinliang has quit IRC | 09:00 | |
*** xinliang has joined #openstack-dib | 09:11 | |
*** hashar has joined #openstack-dib | 09:17 | |
*** openstack has joined #openstack-dib | 09:57 | |
*** hashar has quit IRC | 09:57 | |
*** owalsh_ is now known as owalsh | 09:59 | |
*** hashar has joined #openstack-dib | 10:03 | |
*** makowals has joined #openstack-dib | 10:20 | |
*** chhavi has quit IRC | 10:24 | |
*** hashar has quit IRC | 11:19 | |
*** brault has joined #openstack-dib | 11:31 | |
*** pmannidi has quit IRC | 11:37 | |
*** pmannidi has joined #openstack-dib | 11:37 | |
*** pmannidi has quit IRC | 12:07 | |
*** hashar has joined #openstack-dib | 12:29 | |
*** hashar has quit IRC | 12:35 | |
*** hashar has joined #openstack-dib | 12:35 | |
*** makowals has quit IRC | 12:55 | |
*** makowals has joined #openstack-dib | 12:56 | |
*** makowals has quit IRC | 13:12 | |
*** makowals has joined #openstack-dib | 13:22 | |
*** hashar has quit IRC | 13:40 | |
*** makowals has quit IRC | 14:06 | |
*** makowals has joined #openstack-dib | 14:14 | |
*** hashar has joined #openstack-dib | 14:41 | |
openstackgerrit | Andreas Scheuring proposed openstack/diskimage-builder master: WIP: Add zipl support for s390 architecture with SCSI boot https://review.openstack.org/466308 | 14:46 |
*** hashar has quit IRC | 15:13 | |
*** brault has quit IRC | 16:42 | |
*** sauloaislan has quit IRC | 16:48 | |
*** Sukhdev has joined #openstack-dib | 17:47 | |
*** brault has joined #openstack-dib | 17:59 | |
*** brault has quit IRC | 18:03 | |
*** andreas-f has joined #openstack-dib | 18:16 | |
*** Sukhdev has quit IRC | 19:37 | |
openstackgerrit | Jesse Keating proposed openstack/diskimage-builder master: Remove use of 'which'. https://review.openstack.org/466063 | 19:43 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Switch to using "partition" objects https://review.openstack.org/465916 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Move exception to it's own file (again) https://review.openstack.org/466129 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Move parts of Partition creation into object https://review.openstack.org/465838 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Split partition into it's own file https://review.openstack.org/465837 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Switch debian to deb.debian.org https://review.openstack.org/466182 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Remove tree config parser from plugins https://review.openstack.org/466183 | 20:15 |
openstackgerrit | Andreas Florath proposed openstack/diskimage-builder master: Move to generic configuration parsing https://review.openstack.org/465417 | 20:15 |
*** andreas-f has quit IRC | 20:51 | |
*** andreas-f has joined #openstack-dib | 20:52 | |
*** andreas-f has quit IRC | 20:53 | |
*** Sukhdev has joined #openstack-dib | 21:14 | |
*** yolanda has quit IRC | 21:18 | |
*** yolanda has joined #openstack-dib | 21:20 | |
*** jamielennox is now known as jamielennox|away | 23:00 | |
*** Sukhdev has quit IRC | 23:22 | |
*** jamielennox|away is now known as jamielennox | 23:40 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!