[00:12:16] *** Joins: tomzawadzki (~tzawadzk@134.134.139.78) [00:12:17] *** Joins: Tomasz (~tzawadzk@134.134.139.78) [01:00:21] *** Quits: ziyeyang_ (~ziyeyang@134.134.139.73) (Remote host closed the connection) [01:33:50] *** Parts: tomzawadzki (~tzawadzk@134.134.139.78) ("Leaving") [01:35:38] *** Parts: Tomasz (~tzawadzk@134.134.139.78) ("Leaving") [01:36:02] *** Joins: tomzawadzki (~tzawadzk@134.134.139.78) [02:01:18] *** Quits: changpeng (~changpeng@192.55.54.42) (Remote host closed the connection) [03:25:20] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [04:55:47] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [05:34:57] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [06:42:30] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [07:09:30] *** Quits: tomzawadzki (~tzawadzk@134.134.139.78) (Ping timeout: 240 seconds) [07:27:33] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [07:42:14] *** Quits: Edwinzrodriguez (d8f01e17@gateway/web/freenode/ip.216.240.30.23) (Ping timeout: 260 seconds) [07:43:02] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [07:44:50] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [08:47:32] bwalker: can you check this patch again after my rebase? https://review.gerrithub.io/#/c/362628/ [09:14:25] done [09:14:35] pushed [09:14:54] I'll rebase my series again now that Darek's stuff is merged [09:15:21] 13 patches in the series - 2 failed due to the stub issue, 1 due to hotplug test [09:16:21] i like this one: http://spdk.intel.com/public/spdk/builds/review/02244104542faa83fda72aae1be1ea220025a23b.1495787286/ [09:16:35] how did that get so fast? [09:17:11] isn't that like 30 seconds faster than we expected? [09:17:23] that's using the stub for nvmf [09:17:50] want to merge that one before I rebase? [09:19:59] my only pause is it should have been broken up into a few different patches [09:20:16] yeah at least 3 I think [09:20:25] but it does look right to me [09:20:32] yeah - it all looks good [09:20:40] the nvmf code change specifically i'd like separated out [09:20:51] with an explanation for why it's needed [09:20:55] i'll comment [09:25:06] I rebased my series and I'm going to push it like 2 patches at a time [09:25:24] if we can get them merged - everyone has approved the early ones in the series at some point in time [09:25:38] sounds good [09:27:33] so if we get vhost moved to a new test system, that would reduce fedora-07 down to about 5 minutes flat [09:27:42] after the nvmf stub changes [09:28:10] which would make rocksdb the long pole in the tent [09:29:43] we could cut the database size to 30GB from 50GB and each of the 4 workload test runs from 30s to 20s which would get that down to about 5 minutes flat [09:32:25] yeah and then that's about the best we can do [09:32:57] we don't need to run scan-build on that system [09:33:11] we already disable that on fedora-07 which saves about 20-30 seconds [09:33:25] but yeah, 5 minutes is about the minimum [09:34:04] unless we have multiple rocksdb test systems that each are responsible for testing different workloads [09:34:36] i'd be ecstatic with 5 minutes though [09:51:27] bwalker: i want to come up with a better name than "blobs as bdevs" - what about "blobvols"? how much do you hate that? [09:52:17] it's better tahn blobs as bdevs [09:52:30] that's a pretty low bar [09:52:52] we don't necessarily have to have blob in the name [09:53:09] we could just call them logical volumes [09:54:56] spdk_logical_volume? logical_volume.c? [09:55:05] that's fine with me [09:55:14] everyone knows what that is, and using the blobstore is an implementation detail [09:56:14] ok - let's do logical_vol (save those three characters), and logical_volstore [09:56:19] local_vol => blob [09:56:24] logical_vol => blob [09:56:32] logical_volstore => blobstore [09:57:20] thinking we explicitly create a logical_volstore on a blobstore by creating a superblob to denote it as a volstore (different from a blobfs) [09:57:41] yeah that sounds reasonable [09:58:01] so when the volstore module tastes a new bdev, it will (1) try to spdk_bs_init [09:58:18] and if (1) succeeds, (2) get super blob and check for volstore signature [09:58:40] yeah - will work now that ziyeyang has made the init process async [09:58:42] if those both succeed then it iterates through all of the blobs and registers them as bdevs [09:58:46] exactly [10:00:08] we can get this one in before your next set of patches to save a few more seconds per test run: https://review.gerrithub.io/#/c/362629/ [10:16:21] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [10:18:26] ok merged first two of my series, then your nic one [10:18:33] and I pushed the next 3 in my series rebased [10:21:31] fyi - i removed your +2 from the nvmf stub patch [10:21:43] ok [10:21:51] it hides the -1, I should try to figure that out [11:15:52] bwalker: this is another one that should help on the iscsi test time: https://review.gerrithub.io/#/c/361907/ [11:54:01] https://review.gerrithub.io/#/c/362773/ [11:54:05] coalescing unmaps for clusters [11:54:12] oh, you already reviewed it [11:54:15] :) [14:23:04] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [14:32:11] i added comments to a couple of your i/o channel patches [14:32:58] i'm ok deferring them - we can add a patch at the end that does more checks before freeing threads (i.e. checking that the channel list is empty, and there are no outstanding send_msg calls in flight) [14:41:38] bwalker: i pushed the first 5 patches in your series [15:49:50] *** Quits: sethhowe (~sethhowe@134.134.139.74) (Remote host closed the connection)