Displaying #traffic-server/2015-12-09.log:

Wed Dec 9 00:07:03 2015  dxu:Joined the channel
Wed Dec 9 00:08:35 2015  dxu_:Joined the channel
Wed Dec 9 00:13:46 2015  dxu_:Left the channel
Wed Dec 9 00:25:58 2015  reveller1:Joined the channel
Wed Dec 9 00:42:36 2015  blattj:Joined the channel
Wed Dec 9 00:50:31 2015  biilmann:Joined the channel
Wed Dec 9 00:51:23 2015  danielxu:Joined the channel
Wed Dec 9 01:08:42 2015  biilmann:Joined the channel
Wed Dec 9 02:40:42 2015  _klk_:Joined the channel
Wed Dec 9 02:59:21 2015  danielxu:Joined the channel
Wed Dec 9 03:42:57 2015  biilmann:Joined the channel
Wed Dec 9 03:49:47 2015  reveller2:Joined the channel
Wed Dec 9 04:02:13 2015  es:Joined the channel
Wed Dec 9 04:04:33 2015  shinrich1:Joined the channel
Wed Dec 9 05:17:14 2015  gancho:Joined the channel
Wed Dec 9 05:29:38 2015  danielxu:Joined the channel
Wed Dec 9 07:05:32 2015  mturk:Joined the channel
Wed Dec 9 07:05:32 2015  mturk:Joined the channel
Wed Dec 9 08:11:45 2015  slx86:Joined the channel
Wed Dec 9 08:16:56 2015  slx86:Joined the channel
Wed Dec 9 09:01:06 2015  biilmann:Joined the channel
Wed Dec 9 09:16:59 2015  Lethalman:Joined the channel
Wed Dec 9 10:29:42 2015  bahumbug:Joined the channel
Wed Dec 9 10:29:57 2015  niq:Joined the channel
Wed Dec 9 13:21:06 2015  niq:Joined the channel
Wed Dec 9 13:21:26 2015  bahumbug:Joined the channel
Wed Dec 9 14:17:32 2015  swoc:Joined the channel
Wed Dec 9 14:30:28 2015  esproul:Joined the channel
Wed Dec 9 14:34:06 2015  briang:hey swoc, yamc: TS-4052 looks good, what was your question about it?
Wed Dec 9 14:34:22 2015  yamc:I changed your code and so I thought I'd run it by you.
Wed Dec 9 14:34:43 2015  yamc:I think it's reasonable to presume that where shared_ptr is, there also is unique_prtr.
Wed Dec 9 14:35:39 2015  yamc:Can you take a quick look at TS-3235, particularly the last few comments? There's a claim of a potential issue with the CPP API.
Wed Dec 9 14:37:14 2015  yamc:Basically if your'e calling InterceptPlugin::produce from another (non-ATS) thread there's no way to lock the continuation.
Wed Dec 9 14:37:47 2015  yamc:I think adding the lock to the two methods is reasonable - either you're in an ET_NET thread and so already have the lock, or you're not in which case you really need it.
Wed Dec 9 14:38:01 2015  yamc:I think we're having crashes in Y! because of this.
Wed Dec 9 14:38:15 2015  briang:yamc, you can't do stuff like that
Wed Dec 9 14:38:24 2015  briang:I think you need to call TSThreadInit() from not ATS created threads
Wed Dec 9 14:38:31 2015  briang:if you plan on calling back into stuff like that, right?
Wed Dec 9 14:38:39 2015  briang:or create all your threads with TSThreadCreate() ?
Wed Dec 9 14:38:47 2015  briang:does that sound right?
Wed Dec 9 14:38:56 2015  yamc:yes, but it's a locking problem. There's a continuation (state_->cont_) with a mutex.
Wed Dec 9 14:39:11 2015  yamc:That mutex is different from the state_->mutex_
Wed Dec 9 14:39:22 2015  yamc:InterceptPlugin locks the latter but not the former.
Wed Dec 9 14:39:30 2015  yamc:While ET_NET locks the other way.
Wed Dec 9 14:39:37 2015  briang:so you're deadlocking?
Wed Dec 9 14:39:46 2015  yamc:No, crashing because of race conditions.
Wed Dec 9 14:40:24 2015  briang:ok I can read through it first thing in the morning
Wed Dec 9 14:40:30 2015  briang:It's already pretty late here
Wed Dec 9 14:40:31 2015  yamc:Hmmm, deadlock. Let me think about that.
Wed Dec 9 14:40:33 2015  briang:but I will do it today
Wed Dec 9 14:40:44 2015  yamc:Morning is fine.
Wed Dec 9 14:40:54 2015  briang:any others besides TS-3235?
Wed Dec 9 14:40:58 2015  yamc:My plan is to try it out here internally so see if it (1) works (2) fixes the problem.
Wed Dec 9 14:41:02 2015  yamc:Not currently.
Wed Dec 9 14:41:03 2015  yamc:Thanks.
Wed Dec 9 14:41:22 2015  yamc:If I get the messaging stuff working, I'll want to add that to the CPP API.
Wed Dec 9 14:41:31 2015  yamc:I need to do more refinement first.
Wed Dec 9 14:41:47 2015  briang:So we've had good luck doing things like producing to a transformation from a non-ats thread (initialized via TSThreadInit)
Wed Dec 9 14:42:01 2015  briang:so I think we just need to do what we've done in transformations w/ intercepts
Wed Dec 9 14:42:12 2015  yamc:Ah, good point. I'll look at that.
Wed Dec 9 14:42:21 2015  briang:but I'll dig through the code, to be honest I know transformations well but manjesh wrote all the intercept code
Wed Dec 9 14:42:40 2015  briang:let me know if you see anything before I take a look later
Wed Dec 9 14:42:45 2015  yamc:OK
Wed Dec 9 14:54:28 2015  yamc:Ah, the difference is TransformPlugin doesn't call TSContCreate.
Wed Dec 9 14:54:59 2015  yamc:And it's that continuation that's at the root of the race condition.
Wed Dec 9 14:55:55 2015  yamc:briang^
Wed Dec 9 15:13:56 2015  mturk_:Joined the channel
Wed Dec 9 15:13:56 2015  mturk_:Joined the channel
Wed Dec 9 15:31:56 2015  Zerpex:Joined the channel
Wed Dec 9 15:34:30 2015  shinrich1:Joined the channel
Wed Dec 9 15:52:48 2015  biilmann:Joined the channel
Wed Dec 9 15:57:44 2015  gancho_:Joined the channel
Wed Dec 9 16:00:31 2015  JSeymour:Joined the channel
Wed Dec 9 16:00:33 2015  jumby:Joined the channel
Wed Dec 9 16:02:20 2015  reveller:Joined the channel
Wed Dec 9 16:09:04 2015  reveller:shinrich1: I have time to work on the ssl_cert_loader plugin. Do you have any notes, dox, etc on the plugin?
Wed Dec 9 16:10:04 2015  yamc:reveller - what are you going to do with it? I was working on it as well.
Wed Dec 9 16:11:21 2015  reveller:with the "encrypt All The Things movement" looming just over the horizon, there is going to be a huge influx of new certs that I am going to have to deal with
Wed Dec 9 16:12:07 2015  reveller:I am looking for a method of dynamically configuring certs without having to reload the config
Wed Dec 9 16:12:16 2015  yamc:Hey, that's what I'm doing!
Wed Dec 9 16:12:45 2015  yamc:Stop copying me!
Wed Dec 9 16:13:55 2015  reveller:Right now, I need to replace the ssl_multicert config which is justa simple list of hte certs I need ATS to recognize
Wed Dec 9 16:14:22 2015  yamc:You should talk to Dr. H, she's spent a good bit of time on designing a new config file format for cert loader.
Wed Dec 9 16:14:25 2015  reveller:We should work TOGETHER for the greater good of the Internets! :)
Wed Dec 9 16:14:31 2015  yamc:You should also look at TS-4032.
Wed Dec 9 16:14:35 2015  yamc:That's what I'm working on.
Wed Dec 9 16:14:44 2015  reveller:since her last efforts?
Wed Dec 9 16:14:56 2015  reveller:checking
Wed Dec 9 16:14:57 2015  yamc:I think so. It's been over the last couple of weeks.
Wed Dec 9 16:15:20 2015  yamc:I picked cert loader as my demonstration plugin, to show the utility of this mechanism.
Wed Dec 9 16:16:07 2015  reveller:ah, I see... was a bit confused how command line messaging was related to ssl cert config
Wed Dec 9 16:16:37 2015  yamc:The plan was to show how nice it would be to be able to tweak around the certs from the commandline without reloading.
Wed Dec 9 16:16:53 2015  reveller:nice
Wed Dec 9 16:17:08 2015  yamc:or just check an SNI name - "Generate a log message saying what cert is associated with SNI name X"
Wed Dec 9 16:17:45 2015  yamc:I thought that would be handy for spot checking / validating configurations.
Wed Dec 9 16:17:54 2015  yamc:Ask the live table what it thinks it is doing.
Wed Dec 9 16:18:36 2015  yamc:I've almost got that working, I should try to finish at least that part and push it up as a first example.
Wed Dec 9 16:18:55 2015  reveller:indeed... I could incorporate that into my ssl cert sync daemon... write the cert, update config, via the cmd line ask the if the config is what I thought it should be
Wed Dec 9 16:18:58 2015  yamc:The message passing itself, AFAICT, is fully functional.
Wed Dec 9 16:19:22 2015  yamc:In theory you sohould be able to update single certs in the config as well "Load this cert for me"
Wed Dec 9 16:19:30 2015  yamc:Or break SNI <-> cert associations.
Wed Dec 9 16:19:46 2015  reveller:so what can I do to help Dr H?
Wed Dec 9 16:20:23 2015  yamc:That's a good question. She's off at an internal Y! conference this week.
Wed Dec 9 16:20:53 2015  yamc:Can you teleconference from where you are? I'm sure she'd be OK with that to get organized.
Wed Dec 9 16:21:06 2015  bahumbug:Joined the channel
Wed Dec 9 16:21:17 2015  niq:Joined the channel
Wed Dec 9 16:37:15 2015  reveller:yamc: I can teleconference. Do it every day. Just need to set up a time. I will be traveling on Friday, and will be working in AZ through the holiday.
Wed Dec 9 16:37:35 2015  yamc:Ping Dr. H on Monday or over the weekend. She'll be back in town by Saturday.
Wed Dec 9 16:38:39 2015  reveller:will do... thanks
Wed Dec 9 16:47:46 2015  blattj:Joined the channel
Wed Dec 9 16:47:53 2015  shinrich1:Joined the channel
Wed Dec 9 17:02:11 2015  es:Joined the channel
Wed Dec 9 17:09:27 2015  es1:Joined the channel
Wed Dec 9 17:11:31 2015  reveller::Qa
Wed Dec 9 17:14:24 2015  gancho:Joined the channel
Wed Dec 9 17:32:57 2015  shinrich1:Joined the channel
Wed Dec 9 17:48:41 2015  blattj:Joined the channel
Wed Dec 9 18:01:02 2015  blattj1:Joined the channel
Wed Dec 9 18:03:12 2015  _klk_:Joined the channel
Wed Dec 9 18:08:48 2015  bcall:zwoop: are you getting failures for gmake check on master?
Wed Dec 9 18:24:29 2015  blattj:Joined the channel
Wed Dec 9 18:33:08 2015  gancho:bcall: I run 'gmake check' against branch ts-4023 which was just rebased against master and it worked
Wed Dec 9 18:43:18 2015  swoc:Joined the channel
Wed Dec 9 18:43:37 2015  bcall:gancho: k
Wed Dec 9 18:44:24 2015  gancho:bcall: tried 'make check' master and looked more carefully - get "parser" like "Error: [parseChunk] Failed to parse chunk..."
Wed Dec 9 18:46:29 2015  bcall:gancho: I see:
Wed Dec 9 18:46:29 2015  bcall:FAIL: test_List
Wed Dec 9 18:46:34 2015  bcall:FAIL: test_X509HostnameValidator
Wed Dec 9 18:46:34 2015  bcall:FAIL: test_Map
Wed Dec 9 18:46:53 2015  bcall:running fedora 23
Wed Dec 9 18:50:09 2015  gancho:the errors I see are in plugins/experimental/esi in Debug (parser_test) ...
Wed Dec 9 18:50:24 2015  gancho:don't see the one you mentioned above
Wed Dec 9 18:53:52 2015  swoc:Joined the channel
Wed Dec 9 18:54:13 2015  bcall:gancho: I am running with asan, so it looks like that is the issue
Wed Dec 9 18:54:24 2015  bcall:I will fix the asan errors
Wed Dec 9 18:55:32 2015  gancho:oh, I tested make check on centos6
Wed Dec 9 18:55:42 2015  gancho:ok
Wed Dec 9 18:55:53 2015  gancho:cool
Wed Dec 9 18:58:30 2015  mturk__:Joined the channel
Wed Dec 9 19:37:52 2015  bahumbug:Joined the channel
Wed Dec 9 19:38:05 2015  niq:Joined the channel
Wed Dec 9 19:38:20 2015  Humbedooh:zwoop: ping
Wed Dec 9 19:39:12 2015  Humbedooh:oh, no wait...unping
Wed Dec 9 19:39:43 2015  zwoop:unack
Wed Dec 9 19:40:09 2015  blattj:Joined the channel
Wed Dec 9 19:44:17 2015  biilmann:Joined the channel
Wed Dec 9 19:47:27 2015  _klk_:Joined the channel
Wed Dec 9 19:58:43 2015  igalic:Joined the channel
Wed Dec 9 19:59:10 2015  swoc:Anyone what's breaking the CI?
Wed Dec 9 20:02:26 2015  zwoop:looks like john's new LRU / CFLUS tests
Wed Dec 9 20:03:15 2015  jpeach:those take a while to run ... maybe move them to R 3 level? or increase the jenkins timeout
Wed Dec 9 20:07:48 2015  zwoop:There is no timeout on Jenkins I'm fairly sure
Wed Dec 9 20:07:57 2015  zwoop:I've had regressions that hang run for hours :)
Wed Dec 9 20:10:26 2015  jpeach:do the CI hosts have memory limits that could get the job killed?
Wed Dec 9 20:17:56 2015  zwoop:that's possible, they each have 2GB RAM
Wed Dec 9 20:19:43 2015  zwoop:I can try to run it manually see if it trips
Wed Dec 9 20:26:22 2015  zwoop:weird, all the builds are nuked
Wed Dec 9 20:26:27 2015  zwoop:let me kick off a new one on master
Wed Dec 9 20:38:39 2015  blattj:Joined the channel
Wed Dec 9 20:41:12 2015  zwoop:jpeach it looks like the RAM cache test killed the FreeBSD bot completely
Wed Dec 9 20:59:25 2015  zwoop:jpeach any brilliant ideas what to do for now? Disable these tests ?
Wed Dec 9 20:59:39 2015  zwoop:I had to hardreboot the FreeBSD bot
Wed Dec 9 21:08:02 2015  jpeach:zwoop: maybe change those tests to REGRESSION_TEST_EXTENDED level?
Wed Dec 9 21:26:54 2015  dxu:Joined the channel
Wed Dec 9 21:35:59 2015  blattj:Joined the channel
Wed Dec 9 21:40:29 2015  zwoop:sure
Wed Dec 9 21:40:32 2015  zwoop:make it so #1!
Wed Dec 9 21:43:56 2015  Humbedooh:you sure do change whoever is #1 a lot
Wed Dec 9 21:44:05 2015  Humbedooh:I'm beginning to think it's all a ruse
Wed Dec 9 21:51:44 2015  gancho:Joined the channel
Wed Dec 9 21:52:38 2015  danielxu:Joined the channel
Wed Dec 9 21:56:16 2015  zwoop:Humbedooh you're the real #1, everyone else is faux
Wed Dec 9 21:56:46 2015  Humbedooh:<3
Wed Dec 9 22:01:50 2015  Humbedooh:zwoop: now answer my PM, please :)
Wed Dec 9 22:04:20 2015  biilmann:Joined the channel
Wed Dec 9 22:06:22 2015  gancho:Joined the channel
Wed Dec 9 22:16:33 2015  shinrich1:Joined the channel
Wed Dec 9 22:18:20 2015  shinrich2:Joined the channel
Wed Dec 9 22:42:16 2015  _klk_:Joined the channel
Wed Dec 9 22:51:59 2015  gancho:Joined the channel
Wed Dec 9 23:00:21 2015  zwoop:jpeach quick look at http://ur1.ca/oc47p please ?
Wed Dec 9 23:01:15 2015  jpeach:+1
Wed Dec 9 23:06:28 2015  _klk_:Joined the channel
Wed Dec 9 23:07:08 2015  zwoop:I'm just gonna commit this (using my poor judgement and skipping the PR)
Wed Dec 9 23:07:11 2015  zwoop:I mean, good judgement ....
Wed Dec 9 23:09:14 2015  zwoop:I'm working with the OmniOS guys too to finally get a build-bot setup there.
Wed Dec 9 23:33:27 2015  gancho:Joined the channel
Wed Dec 9 23:33:35 2015  blattj:Joined the channel
Wed Dec 9 23:48:48 2015  blattj1:Joined the channel
Wed Dec 9 23:58:19 2015  gancho:Joined the channel

Comments