Thursday, 2016-09-22

*** apearson has quit IRC00:02
*** smatzek has quit IRC00:04
*** thorst has joined #openstack-powervm00:59
*** thorst has quit IRC01:04
*** dwayne has quit IRC01:23
*** dwayne has joined #openstack-powervm01:24
*** thorst has joined #openstack-powervm01:45
*** thorst has quit IRC01:50
*** thorst has joined #openstack-powervm02:17
*** dwayne has quit IRC02:21
*** thorst has quit IRC02:21
*** thorst has joined #openstack-powervm02:48
*** thorst has quit IRC02:53
*** jwcroppe has quit IRC03:07
*** jwcroppe has joined #openstack-powervm03:08
*** kotra03 has joined #openstack-powervm03:19
*** mdrabe_ has quit IRC03:54
*** mdrabe_ has joined #openstack-powervm03:55
*** mdrabe_ has quit IRC04:31
*** thorst has joined #openstack-powervm04:48
*** kotra03 has quit IRC04:49
*** apearson has joined #openstack-powervm05:16
*** thorst has quit IRC05:42
*** thorst has joined #openstack-powervm05:45
*** thorst has quit IRC05:54
*** kotra03 has joined #openstack-powervm06:02
*** apearson has quit IRC06:13
*** thorst has joined #openstack-powervm06:24
*** thorst has quit IRC06:30
*** thorst has joined #openstack-powervm06:41
*** AlexeyAbashkin has quit IRC06:44
*** AlexeyAbashkin has joined #openstack-powervm06:46
*** thorst has quit IRC06:46
*** thorst has joined #openstack-powervm06:57
*** thorst has quit IRC07:05
*** jwcroppe has quit IRC07:32
*** thorst has joined #openstack-powervm07:33
*** jwcroppe has joined #openstack-powervm07:36
*** thorst has quit IRC07:41
*** thorst has joined #openstack-powervm08:11
*** thorst has quit IRC08:16
*** thorst has joined #openstack-powervm08:28
*** thorst has quit IRC08:33
*** k0da has joined #openstack-powervm08:38
*** thorst has joined #openstack-powervm08:44
*** thorst has quit IRC08:51
*** kotra03 has quit IRC08:56
*** kotra03 has joined #openstack-powervm08:58
*** thorst has joined #openstack-powervm09:23
*** thorst has quit IRC09:28
*** thorst has joined #openstack-powervm09:40
*** thorst has quit IRC09:44
*** thorst has joined #openstack-powervm09:56
*** thorst has quit IRC10:04
*** viclarson has joined #openstack-powervm10:10
*** thorst has joined #openstack-powervm10:16
*** thorst has quit IRC10:20
*** thorst has joined #openstack-powervm10:32
*** kotra03 has quit IRC10:34
*** thorst has quit IRC10:40
*** thorst has joined #openstack-powervm11:08
*** smatzek has joined #openstack-powervm11:09
*** thorst has quit IRC11:16
*** viclarson_ has joined #openstack-powervm11:26
*** viclarson has quit IRC11:30
*** kotra03 has joined #openstack-powervm11:31
*** thorst has joined #openstack-powervm11:43
*** thorst has quit IRC11:52
*** edmondsw has joined #openstack-powervm12:02
*** thorst has joined #openstack-powervm12:19
*** thorst has quit IRC12:28
*** mdrabe has joined #openstack-powervm12:29
*** dwayne has joined #openstack-powervm12:34
*** tblakes has joined #openstack-powervm12:50
*** thorst_ has joined #openstack-powervm12:59
*** efried has joined #openstack-powervm12:59
efriedviclarson_, we keep missing each other.13:00
efriedapearson said the following yesterday:13:00
efried"efried / viclarson_ - so the first thing to check - are the VIOSs up and running?   Assuming so, grab the console for the novalink lpar.  Where is it stopping?  Is it sitting at SMS?  If so, try selecting the boot device and continue the boot.  Is it erroring out on boot due to a filesystem that needs to be fsck'ed?  If so, choose Recovery from the grub prompt and boot into the recovery mode to run fsck on the filesystem.13:00
*** thorst_ has quit IRC13:04
viclarson_i think it were unrecoverable errors at filesystem13:09
efriedOkay.  Let us know if you have further issues.13:09
viclarson_ty, doensnt matter, already reinstalled13:09
efriedAnd it's working now?13:09
viclarson_looks like13:10
efriedGreat13:10
viclarson_reinstalled from scratch13:10
viclarson_not with recovery13:10
viclarson_recovery doenst work13:10
efriedSome problems are... unrecoverable.13:10
viclarson_mb13:11
efriedI'm curious (because we like to know who's using our stuff) - where are you from, and what's your interest in NovaLink?13:11
efriedadreznec, yt?13:13
efriedCan you please +2 https://review.openstack.org/#/c/373415/ (since I've "contributed" to it now)?13:13
*** thorst_ has joined #openstack-powervm13:35
*** smatzek has quit IRC13:36
*** thorst_ has quit IRC13:40
*** viclarson_ has quit IRC13:43
*** thorst_ has joined #openstack-powervm13:51
*** burgerk has joined #openstack-powervm13:56
*** thorst_ has quit IRC13:56
*** burgerk has quit IRC14:07
*** thorst_ has joined #openstack-powervm14:07
*** dwayne has quit IRC14:11
*** tjakobs has joined #openstack-powervm14:14
*** thorst_ has quit IRC14:16
adreznecefried: sure, I'll take a look here14:26
efriedThanks14:26
efriedadreznec, it was +2 - all I changed was the path to the pypowervm req14:26
*** thorst_ has joined #openstack-powervm14:29
adreznecefried: yep, LGTM. Looks ready to merge as well, I'll throw a WF+1 on it here14:29
efriedOr I can.  Either way.14:30
*** tblakes has quit IRC14:30
efriedchmod666org is waiting for it ;-)14:30
adreznecIt better merge soon then, otherwise he could find Drew and I here at Edge and hold us hostage until it does!14:32
*** smatzek has joined #openstack-powervm14:37
*** smatzek has quit IRC14:38
*** smatzek_ has joined #openstack-powervm14:38
*** smatzek_ is now known as smatzek14:39
*** burgerk has joined #openstack-powervm14:39
*** burgerk_ has joined #openstack-powervm14:41
*** burgerk has quit IRC14:44
*** tblakes has joined #openstack-powervm14:45
efriedthorst_, any chance of getting 3979 reviewed?14:45
thorst_will try...I need about 30 minutes for other stuff :-(14:46
efriedpah, "other stuff"14:46
efriedthorst_, if it helps, the delta from PS2 is almost exclusively doc14:47
thorst_well, I'm just bottlenecked on a final presentation, 100 e-mails and bleh.  Will loop back on it today for sure...just hoping that's in 30 min and not a few hours14:47
*** mdrabe_ has joined #openstack-powervm14:48
*** catintheroof has joined #openstack-powervm14:50
*** mdrabe has quit IRC14:52
*** mdrabe_ is now known as mdrabe14:55
*** viclarson has joined #openstack-powervm15:00
*** dwayne has joined #openstack-powervm15:03
*** viclarson has quit IRC15:08
*** k0da has quit IRC15:29
thorst_efried: what would you think of having two new event types.  One for pypowervm and one for true clients?  Or does that drive change down into the REST API itself?15:35
efriedthorst_, yeah, REST API controls the type entirely.  When we (any REST client) PUT one, it's CUSTOM_CLIENT_EVENT.15:35
thorst_k...15:36
thorst_bleh15:36
efriedIt's based on an enum, so at the very least, they can't just let us arbitrarily use any string.15:36
thorst_right right15:36
efriedWe could ask for a new enum15:36
efriedBut how do we "prove" we're coming from pypowervm?15:36
thorst_no, because they shouldn't know about pypowervm as a client15:36
efriedright15:36
thorst_right...exactly15:36
efriedWe need to use the data/detail to differentiate.15:36
thorst_but its kinda where I'm going with how does neutron know that pypowervm is producing an event.15:37
thorst_but yeah, like you said15:37
efriedAlong the lines of what I did in https://review.openstack.org/37353215:37
thorst_that can be done in data/detail.15:37
thorst_gotta run...review looks OK but I'm not through yet15:37
thorst_will be back in 3015:37
efriedk15:37
*** thorst_ has quit IRC15:37
*** thorst_ has joined #openstack-powervm16:08
*** jwcroppe has quit IRC16:10
*** thorst_ has quit IRC16:12
*** thorst_ has joined #openstack-powervm16:18
*** thorst__ has joined #openstack-powervm16:20
*** thorst_ has quit IRC16:22
*** viclarson has joined #openstack-powervm16:38
*** seroyer has joined #openstack-powervm16:40
*** k0da has joined #openstack-powervm17:01
*** thorst__ has quit IRC17:11
*** seroyer has quit IRC17:12
*** apearson has joined #openstack-powervm17:12
*** thorst_ has joined #openstack-powervm17:14
*** thorst__ has joined #openstack-powervm17:16
*** esberglu has joined #openstack-powervm17:16
*** thorst_ has quit IRC17:18
*** thorst__ has quit IRC17:21
*** thorst_ has joined #openstack-powervm17:21
*** viclarson has quit IRC17:24
*** thorst_ has quit IRC17:25
*** apearson has quit IRC17:33
*** apearson has joined #openstack-powervm17:36
*** apearson_ has joined #openstack-powervm17:41
*** apearson has quit IRC17:41
*** thorst_ has joined #openstack-powervm17:51
*** thorst__ has joined #openstack-powervm17:56
*** thorst_ has quit IRC17:56
*** thorst__ has quit IRC17:57
*** thorst_ has joined #openstack-powervm17:57
*** seroyer has joined #openstack-powervm17:58
*** thorst_ has quit IRC18:24
*** thorst_ has joined #openstack-powervm18:25
*** seroyer has quit IRC18:28
*** apearson_ has quit IRC18:29
*** thorst_ has quit IRC18:30
*** thorst_ has joined #openstack-powervm18:31
efriedthorst_, what do you think of eliminating the interval altogether?18:31
efried(edmondsw ^^)18:31
efriedThe fact that it's there at all implies that the author (understandably - no docs and a weird API) didn't know about the pseudo-synchronous business.   The average consumer will realize a "performance boost" if we get rid of it.  And I can't see a downside.  It's not like it will cause significant extra thrashing.  In the worst case - a listener on a totally steady-state system - you'll issue a GET every 10s instead of eve18:33
edmondswi was the author, and I'm still not aware of what you mean by pseudo-synchronous18:35
*** thorst_ has quit IRC18:35
efriededmondsw, just so.  Here's how it works:18:38
efriedWhen you do the GET, if the REST server has any events waiting, it returns 'em all right away.18:38
efriedIf there are no events waiting, it waits for some to arrive; and as soon as any do, it returns 'em all right away.18:38
efriedIf it has waited for 10s, it returns an empty feed.18:38
efriedSo it's designed explicitly for the client to poll it in a hard loop, with no client-side sleeping, for optimal efficiency.18:39
efriedThat is, you'll get events as soon as they're available.  You don't have to worry about how much stuff happened while you were sleeping.18:39
efriedAnd yes, it's very clever.  The author applied for a patent.  (Though as far as I can tell, it has not been issued yet.)18:39
*** kotra03 has quit IRC18:47
*** esberglu has quit IRC19:03
*** thorst_ has joined #openstack-powervm19:08
edmondswefried, so if it waited for 10s, and returns an empty feed, we'd just immediately call again?19:10
efriedyup19:10
edmondswcool19:10
mdrabethat's patentable?19:10
efriedAnd it would wait another 10s, unless there were events to return.19:10
edmondswyeah, if there isn't already a patent for it19:11
efriedWell, I sure thought so.  The author was skeptical, but I talked him into submitting a disclosure.19:11
efriedAt least, I think he submitted it.  He wrote it up.  Maybe he never pushed it through.19:11
efriedCause I asked the patent folks about it, and they couldn't find it.19:11
edmondswI think there may be a similar web technology19:11
efriedAt this point it's been in the field for long enough that it's moot anyway.19:12
efriedBut it would've been neat if the "similar web technology" came afterward.  $$19:12
edmondswno, I think it's probably older... though maybe different enough... anyway19:13
edmondswso I think I'd be fine with getting rid of the interval altogether and just always doing this19:13
efriedCool.19:13
efriedMaybe I'll propose that as a separate patch.19:13
*** thorst_ has quit IRC19:28
*** miltonm_ has quit IRC19:43
*** thorst_ has joined #openstack-powervm19:54
*** miltonm has joined #openstack-powervm19:57
*** seroyer has joined #openstack-powervm19:58
*** seroyer has quit IRC20:05
*** thorst_ has quit IRC20:06
*** thorst_ has joined #openstack-powervm20:08
*** seroyer has joined #openstack-powervm20:15
*** thorst_ has quit IRC20:16
*** jwcroppe has joined #openstack-powervm20:27
*** seroyer_ has joined #openstack-powervm20:28
*** seroyer has quit IRC20:29
*** seroyer_ is now known as seroyer20:29
*** dwayne has quit IRC20:31
*** smatzek has quit IRC20:32
*** esberglu has joined #openstack-powervm20:37
*** dwayne has joined #openstack-powervm20:54
*** smatzek has joined #openstack-powervm20:55
*** smatzek has quit IRC21:00
*** seroyer has quit IRC21:03
*** esberglu has quit IRC21:13
*** thorst_ has joined #openstack-powervm21:22
*** burgerk__ has joined #openstack-powervm21:24
*** thorst___ has joined #openstack-powervm21:25
*** thorst_ has quit IRC21:27
*** burgerk_ has quit IRC21:28
*** burgerk__ has quit IRC21:29
*** thorst_ has joined #openstack-powervm21:29
*** thorst___ has quit IRC21:30
*** thorst_ has quit IRC21:33
*** edmondsw has quit IRC21:37
*** AlexeyAbashkin has quit IRC21:37
*** miltonm has quit IRC21:37
*** chmod666org has quit IRC21:37
*** k0da has quit IRC21:37
*** adreznec has quit IRC21:37
*** toan has quit IRC21:37
*** AlexeyAbashkin has joined #openstack-powervm21:38
*** k0da has joined #openstack-powervm21:40
*** adreznec has joined #openstack-powervm21:40
*** toan has joined #openstack-powervm21:40
*** AndyWojo has quit IRC21:41
*** edmondsw has joined #openstack-powervm21:41
*** thorst_ has joined #openstack-powervm21:41
*** miltonm has joined #openstack-powervm21:42
*** chmod666org has joined #openstack-powervm21:42
*** thorst_ has quit IRC21:47
*** jwcroppe has quit IRC21:49
*** thorst_ has joined #openstack-powervm21:56
*** thorst__ has joined #openstack-powervm22:00
*** thorst_ has quit IRC22:00
*** dwayne has quit IRC22:01
*** tjakobs has quit IRC22:06
thorst__efried: so for 4010...if we do it in the wrapper22:10
thorst__and then add it later into the REST22:10
thorst__would we then remove it from pypowervm22:10
thorst__resulting in a different signature?22:10
efriedThe REST event would have a different shape.22:10
thorst__and the consumer would need to understand that.22:10
efriedAt the very least, its 'type' would be in the standard ADD_URI/MODIFY_URI/etc.22:11
thorst__right.22:11
*** thorst__ is now known as thorst22:11
efriedSo the consumer would have to understand that anyway.22:11
thorstbleh bleh bleh22:11
*** AndyWojo has joined #openstack-powervm22:11
efriedBut the likelihood of convincing the REST folks to include the MAC address in the event is pretty minimal22:11
efriedSo I would still want to keep the custom event for those.22:11
thorstbut the reason you want a mac is because of the OS code.22:12
efriedThe consumer could switch to the "official" events if they wanted, but they would have to do a real GET.22:12
thorsthonestly, what you could do if you put this in REST is extra detail.  OpenStack details like UUIDs or what not.22:12
thorstbrb22:12
efriedthorst Well, okay, but I'm pretty comfortable with the idea of event detail containing a subset (the "interesting bits") of the object affected.22:12
efriedIt's not like I put in the neutron port UUID (not that I could anyway)22:14
*** mdrabe has quit IRC22:15
*** thorst has quit IRC22:15
*** jwcroppe has joined #openstack-powervm22:16
efriedToday, the 'detail' field contains *almost* that ilk of information, for *some* events.  Like you can get an LPAR mod event where the detail contains the key 'NVRAM', implying that the thing that changed about the LPAR was its NVRAM, and you should only refetch the LPAR if NVRAM is a thing you care about.  Understandably, the detail field doesn't contain the whole NVRAM.  But another example is PartitionState - where they22:16
*** thorst has joined #openstack-powervm22:17
*** thorst has quit IRC22:18
*** thorst has joined #openstack-powervm22:20
thorstsorry - spotty connection22:21
*** tblakes has quit IRC22:23
thorstI guess that I'm just coming from the perspective of...lets forget that the system sends events.  If we can just have nova pass an event to neutron...we don't even really need the port_update in the neutron code itself.22:25
*** apearson has joined #openstack-powervm22:25
efriedIf the events were 100% reliable, yeah.  But we still need to account for "missed events".22:26
thorstare events not reliable?22:26
efriedThings like if the server restarts with events still in the queue that we haven't consumed yet; or we let too much time go by without pulling events; or the agent restarts and has to (re)register the (new) listener.22:27
thorst(we could even use that for sea basically - and I know we can your current approach too)22:27
efriedYeah, I'm rewriting the whole shebang.22:28
thorstserver restarts with events...that's going to be massively disruptive anyway22:28
thorsttoo much time go without pulling events...could that happen?22:28
thorstwhere events are dropped?22:28
thorstor is that just consumer code?22:28
*** jwcroppe has quit IRC22:29
efriedYes, it can happen; I've recreated it with a local test.  I register a listener in ipython, pull some events for a while, then sit there for a while.22:30
efriedNext time I do a pull, I get NEW_CLIENT22:30
thorstbut if we're on a tight two second loop?22:30
efriedTo be clear, the server is really good about letting you know if you need to do a full refetch.22:30
efried(It'll actually be a no-second loop.)  I don't know what criteria the server uses to decide it's time to throw away events.22:31
thorstyeah, I guess I'm just really liking the idea of the nova/neutron having this message queue private to the system to pass events between each other22:31
thorstand not really digging embedding into the wrapper itself.  And the only reason I don't is because it is one off and solely for the purpose of nova/neutron22:31
thorstand once its there...its there forever.22:31
efriedIt might be some number of events - in which case even in a tight loop, if the system is really busy with stuff, it's possible it overloads.22:31
*** seroyer has joined #openstack-powervm22:31
thorstyou don't pull stuff out of pypowervm22:32
efriedI don't agree it's a one-off for the community.  That's just the first place we happen to be using it.22:32
efriedAnd no, I would never *want* to pull it out of pypowervm.22:32
efriedBut like any other event, it only gets consumed by consumers who care about it.22:32
*** jwcroppe has joined #openstack-powervm22:32
thorstright right...22:32
efriedToday the event looper in the agent ignores lots and lots of events.22:32
thorstI guess I'll +1 it, but I'd like to get a second opinion here22:32
efriedIt also happens to consume and act upon a bunch it shouldn't ;-)22:33
efried...which I'm fixing regardless.22:33
*** tjakobs has joined #openstack-powervm22:33
efriedOkay; who would you like to weigh in?22:33
thorstmdrabe and/or adreznec22:33
thorstthis is weird code to anyone though22:33
efriedOkay.  Maybe I send an email.22:34
thorstI'll put my +1 and thought process in here.22:34
efriedSounds good.22:34
*** jwcroppe has quit IRC22:39
*** jwcroppe has joined #openstack-powervm22:46
*** apearson_ has joined #openstack-powervm22:49
*** apearson has quit IRC22:52
*** jwcroppe has quit IRC22:53
*** dwayne has joined #openstack-powervm22:58
*** seroyer has quit IRC23:00
thorstefried: I think https://review.openstack.org/#/c/374439/ may need to go back to stable/mitaka?23:00
thorstmaybe?23:00
openstackgerritMerged openstack/networking-powervm: Switch hacking.checks from neutron to neutron_lib  https://review.openstack.org/37443923:05
openstackgerritMerged openstack/networking-powervm: Remove remaining references to PVID looper  https://review.openstack.org/37436523:06
*** k0da has quit IRC23:18
*** thorst has quit IRC23:19
*** thorst has joined #openstack-powervm23:20
*** apearson_ has quit IRC23:20
*** thorst_ has joined #openstack-powervm23:24
*** thorst has quit IRC23:25
*** thorst has joined #openstack-powervm23:28
*** thorst_ has quit IRC23:29
*** thorst has quit IRC23:32
*** jwcroppe has joined #openstack-powervm23:33
*** jwcroppe_ has joined #openstack-powervm23:35
*** tjakobs has quit IRC23:36
*** jwcroppe has quit IRC23:38
*** thorst has joined #openstack-powervm23:51
*** thorst has quit IRC23:55

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!