[06:34:25] jimharris: I rebased Shuhei's patches (hope he don't mind). Also addressed you comments. Can you take look? https://review.gerrithub.io/#/c/spdk/spdk/+/406919/ [07:51:27] FYI Community meeting in 10 min... [08:27:13] Thanks everyone for attending and for your continued participation in the SPDK Community!! [09:59:09] jimharris: what are your thoughts on the discussion happening here? https://review.gerrithub.io/#/c/399367/ [09:59:16] I'm not sure what the behavior should be [10:12:56] i think this is a case of "you need to know what you're doing" [10:13:14] i'm ok with checking up front if there's enough clusters available [10:14:14] but if you decide to inflate a bunch of blobs in parallel, and you get part way through and find you now don't have enough clusters, then you just fail with -ENOSPC part way through with some of the clusters already allocated [10:14:56] freezing the whole blobstore would be the only way to make this sort-of-atomic and that's a non-starter in my opinion [10:19:10] ok [10:19:19] I'm not sure we should even bother checking [10:19:30] since it isn't reliable [10:24:57] can you take a look at this series when you get a chance? https://review.gerrithub.io/#/c/spdk/spdk/+/409343/ [10:24:59] the first 3 [11:17:39] *** Joins: travis-ci (~travis-ci@ec2-54-92-229-93.compute-1.amazonaws.com) [11:17:40] (spdk/master) test/vhost: remove unused virt-manager config file (Daniel Verkamp) [11:17:40] Diff URL: https://github.com/spdk/spdk/compare/408bb168acc3...8e17d9f21f3c [11:17:40] *** Parts: travis-ci (~travis-ci@ec2-54-92-229-93.compute-1.amazonaws.com) () [11:23:00] i guess i don't think it's terrible to just bail really early if you know there aren't enough clusters before you even start - but we should at least add a comment explaining that the check is not foolproof [12:19:48] very little software if actually foolproof if you think about it... [12:20:43] * peluse is a fool and has proved it many times over :) [12:57:01] bwalker: i reviewed those first three patches [15:26:14] *** Joins: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) [16:30:50] I suggested Young Tack to use the upper case -W and -B for pci-whitelist and pci-blacklist. [16:31:34] This was wrong. I'll fix them first, i.e. -W and -B to -w and -b respectively and I'll prepare another character to delay_subsystem_init. [16:51:51] jim: I posted https://review.gerrithub.io/#/c/spdk/spdk/+/409610/ first. I don't have good idea about the character of delay_subsystem_init. I will use -y instead. [16:53:59] soon we will run out of ascii characters ;) [16:56:23] Yes we have to use better idea but is it OK to use getopt() for now at least in this week? [16:57:29] getopt() is case insensitive in C? [16:57:46] As long as I used in my environment [16:59:43] it isn't in my environment [16:59:49] examples/nvme/perf/perf -Q 10 [17:00:06] invalid option -- 'Q' [17:00:18] OK, I may be wrong please wait for a minute. [17:07:36] jim: sorry getopt() was case-sensitive. The reason why I misunderstood was iscsi-tgt has the option -b (lower-case) and when I specified -b a few functions for pci-whitelist/blacklist were called unexpectedly. [17:08:32] So please forget my mistake. [17:10:35] But I'll check why some functions of -B were called when I specified -b. [17:11:56] jim: By the way, would you check https://review.gerrithub.io/#/c/spdk/spdk/+/406919/ if you can? [17:14:43] of course tomorrow or later is OK for me. [17:19:59] jim: I removed a colon from Pawel's update and all worked as expected. -b is the option to run iscsi-tgt as background and this might cause further confusion to me. [17:28:04] jim: I found a few issues still exist so tomorrow or later may be better for you to review again. [17:45:51] jim: just a notification, now the state of https://review.gerrithub.io/#/c/spdk/spdk/+/406919/ is OK to review again. So, please take a look again when you can. [17:53:47] *** Quits: Shuhei (caf6fc61@gateway/web/freenode/ip.202.246.252.97) (Quit: Page closed)