*** x00350071 has quit IRC | 00:54 | |
*** x00350071 has joined #openstack-smaug | 00:59 | |
*** x00350071_ has joined #openstack-smaug | 03:41 | |
*** x00350071 has quit IRC | 03:44 | |
*** openstackgerrit has quit IRC | 06:17 | |
*** openstackgerrit_ is now known as openstackgerrit | 06:17 | |
*** openstackgerrit_ has joined #openstack-smaug | 06:18 | |
*** openstackgerrit_ is now known as openstackgerrit | 06:18 | |
*** openstackgerrit_ has joined #openstack-smaug | 06:19 | |
yinwei_computer | ping saggi | 07:18 |
---|---|---|
*** yuval has joined #openstack-smaug | 07:27 | |
yuval | ping chenzeng | 07:29 |
saggi | yinwei_computer, yinweiishere: pong | 07:41 |
yinwei_computer | hello saggi | 07:45 |
saggi | yinwei_computer: How are you? | 07:45 |
yinwei_computer | long time not talk to you | 07:45 |
yinwei_computer | good | 07:45 |
yinwei_computer | what about you? | 07:45 |
saggi | I'm good, thank you | 07:46 |
yinwei_computer | hope you will still work on smaug even if you get your job | 07:46 |
yinwei_computer | :) | 07:46 |
yinwei_computer | just want to confirm with the checkpoint collection patch | 07:46 |
yinwei_computer | what's your plan? | 07:47 |
saggi | yinwei_computer: What job? | 07:47 |
yinwei_computer | It's said that you will graduate from school, and you are looking for nice offers | 07:48 |
yinwei_computer | the news is out of date? | 07:48 |
saggi | yinwei_computer: I'm graduating in 6 months but I'm not looking for anything else at the moment. I'll just start working full time. | 07:50 |
yinwei_computer | oh, that's terrific! | 07:50 |
yinwei_computer | I'd like to check the checkpoint patch with you, are you free now? | 07:51 |
saggi | yinwei_computer: Could you send the restore presentation to gample and I. | 07:51 |
yinwei_computer | what restore presentation? the ppt I talked about last week? | 07:52 |
saggi | yes | 07:52 |
yinwei_computer | np | 07:52 |
yinwei_computer | what's your mail box? | 07:52 |
saggi | saggi.miztahi@toganetworks.com | 07:53 |
saggi | saggi.mizrahi@toganetworks.com | 07:54 |
yinwei_computer | got it | 07:55 |
yinwei_computer | I will send it to the mailbox above and eran's huawei mail box | 07:57 |
saggi | yinwei_computer: about checkpoints | 07:57 |
yinwei_computer | yes | 07:58 |
saggi | I sent a very basic patch. After that there are 2 main tasks. Using the leases and serializing the plan. | 07:58 |
yinwei_computer | I see. | 07:59 |
saggi | Can you take one? I think using the lease is perfect for you since you know the alogrithm very well. | 07:59 |
saggi | Since, you helped design it | 08:00 |
yinwei_computer | sure, np | 08:00 |
*** gampel1 has joined #openstack-smaug | 08:00 | |
yinwei_computer | so you will first commit the basic one? | 08:00 |
saggi | Yes, so we can work on top of it together. | 08:00 |
yinwei_computer | and we will work based on your patch? | 08:00 |
saggi | yes | 08:00 |
yinwei_computer | that's good | 08:00 |
saggi | I will do the plan serialization to the checkpoint since I already wrote the tree implementation so I know it well. | 08:00 |
yinwei_computer | another issue I want to confirm with you is | 08:01 |
yinwei_computer | what interface do you think the protection plugin could use to save the information, like backup id, image url stored in bank, or anything else required in save_info_schema? | 08:02 |
yinwei_computer | those information should be saved into checkpoint | 08:02 |
yinwei_computer | per resource | 08:03 |
yinwei_computer | and should be read from checkpoint during restoration | 08:03 |
saggi | The checkpoint file will contain the plan and full resource tree. This means that we hall all the IDs. | 08:05 |
saggi | Each plugin will get a BankSection at /plugindata/<checkpoint_id>/<resiource_id> where it will write whatever it wants. | 08:05 |
saggi | When we get the saved info we will give that section to each plugin and let it read the data and return the dict. | 08:05 |
saggi | The checkpoint file is like and index to all the rest of the data which will be spread out. This will allow multiple plugins to modify data without needing to synchronize their writes since the use different "namespaces" | 08:06 |
yinwei_computer | do you mean in this way, each protection plugin will write directly to bank under the key /plugindata/<checkpoint_id>/<resiource_id> ? they don't need first update this info to checkpoint and then commit checkpoint? | 08:08 |
saggi | It's not a key, it's a prefix. This is what the BankSection is for. | 08:08 |
saggi | It could have any number of objects but will all have the same prefix. | 08:09 |
yinwei_computer | ok, got it. what about read? checkpoint won't load those info from bank to memory, but each protection plugin will read directly from bank by them selves? | 08:10 |
yinwei_computer | and each plugin is free to organize keys under this prefix? | 08:10 |
saggi | They get a BankSection which means they always treat everything as root. | 08:11 |
yinwei_computer | last question is how to get this prefix or how to get this bank section? | 08:12 |
saggi | when the pluing will write to key "/volumeid" it will actually write to /plugindata/<checkpoint_id>/<resource_id>/volumeid. | 08:12 |
saggi | Before passing the context to the plugin. The provider puts the appropriate BankSection in the context. | 08:12 |
saggi | context.bank_section = BankSection(self._bank, "/plugin_data/%s/%s" % (checkpoint_id, resource_id)) | 08:13 |
saggi | The provider will also set it for RO\RW | 08:13 |
saggi | That way the plugin doesn't know about namespaces. | 08:14 |
saggi | Only the provider | 08:14 |
yinwei_computer | hmm, I see | 08:14 |
saggi | When the plugins does context.bank_section.list_objects() it will get all the objects in the namespace. | 08:14 |
saggi | Since this is how the section works | 08:15 |
saggi | It prefixes everything | 08:15 |
saggi | So the plugin can do whatever it wants. | 08:15 |
yinwei_computer | what about to have checkpoint object to return the bank section of each resource? like a static method of checkpoint? | 08:15 |
saggi | Sure, that's good too. | 08:16 |
saggi | It will be better since it already knows the checkpoint id. | 08:16 |
yinwei_computer | ok | 08:16 |
yinwei_computer | yes | 08:16 |
yinwei_computer | another issue I want to confirm with you is the resource graph | 08:16 |
yinwei_computer | do you mean the plan will also persist the resource graph? | 08:17 |
yinwei_computer | just to make sure what we persisted is the actual dependency resource graph instead of a set of top resources | 08:17 |
saggi | No, the checkpoint will. And it will be the entire resource graph. | 08:18 |
yinwei_computer | ok, then I have no problem here. | 08:18 |
saggi | yinwei_computer: :) | 08:19 |
yinwei_computer | so your basic patch is ready, right? | 08:19 |
yinwei_computer | then I will give a +1, so you can merge. and then we can work upon it. | 08:19 |
yinwei_computer | BTW, luobin has a patch about graph walker listener, are you free to review it? | 08:22 |
yinwei_computer | saggi: https://review.openstack.org/#/c/287572/ | 08:24 |
saggi | yinwei_computer: it is ready | 08:25 |
saggi | yinwei_computer: I will look at the patch | 08:25 |
yinwei_computer | saggi, Liu an has also committed a patch, which made a common task flow mechanism for task builder: https://review.openstack.org/#/c/285149/ | 08:26 |
yinwei_computer | hope you can take a look at these two patches when you get time | 08:26 |
yinwei_computer | then we can start debugging protection ASAP | 08:26 |
*** gampel2 has joined #openstack-smaug | 08:54 | |
*** gampel1 has quit IRC | 08:56 | |
*** yinweiishere has quit IRC | 08:59 | |
*** CrayZee has quit IRC | 09:28 | |
*** CrayZee has joined #openstack-smaug | 09:29 | |
*** x00350071_ has quit IRC | 10:22 | |
*** x00350071 has joined #openstack-smaug | 10:24 | |
*** gampel1 has joined #openstack-smaug | 11:58 | |
*** gampel2 has quit IRC | 12:00 | |
*** zengyingzhe has quit IRC | 12:06 | |
openstackgerrit | Yingzhe Zeng proposed openstack/smaug: Implement the ProtectableRegistry https://review.openstack.org/281783 | 12:16 |
*** gampel2 has joined #openstack-smaug | 12:30 | |
*** gampel1 has quit IRC | 12:32 | |
*** x00350071 is now known as xiangxinyong | 14:54 | |
openstackgerrit | Merged openstack/smaug-dashboard: Add skeleton for the smaug-dashboard https://review.openstack.org/289724 | 15:52 |
openstackgerrit | Merged openstack/smaug: Implement the common logic for service client https://review.openstack.org/291528 | 16:00 |
*** yuval has quit IRC | 16:03 | |
openstackgerrit | Merged openstack/smaug: Implement Glance service client https://review.openstack.org/292119 | 16:11 |
openstackgerrit | Yuval Brik proposed openstack/smaug: Implement Nova service client https://review.openstack.org/291532 | 16:26 |
openstackgerrit | Yuval Brik proposed openstack/smaug: Implement Cinder service client https://review.openstack.org/291535 | 16:29 |
*** gampel2 has quit IRC | 16:33 | |
openstackgerrit | Yuval Brik proposed openstack/smaug: Implement Cinder service client https://review.openstack.org/291535 | 16:41 |
*** gampel1 has joined #openstack-smaug | 18:50 | |
*** gampel1 has quit IRC | 20:28 | |
*** gampel1 has joined #openstack-smaug | 20:28 | |
*** gampel1 has quit IRC | 22:28 | |
*** openstackgerrit has quit IRC | 23:03 | |
*** openstackgerrit_ is now known as openstackgerrit | 23:03 | |
*** openstackgerrit has quit IRC | 23:03 | |
*** openstackgerrit_ has joined #openstack-smaug | 23:03 | |
*** openstackgerrit_ is now known as openstackgerrit | 23:04 | |
*** openstackgerrit_ has joined #openstack-smaug | 23:08 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!