| flwang | as for 1535811 | 00:12 |
|---|---|---|
| flwang | i'm wondering if you can get the message id | 00:12 |
| Eva-i | flwang: oh | 00:12 |
| flwang | because at that stage, the message hasn't been written into db | 00:12 |
| flwang | and i don't think it's a bug | 00:14 |
| Eva-i | flwang: you're right, this complicates things much, I'll try to find a solution | 00:14 |
| flwang | it's most like a small feature | 00:14 |
| flwang | don't care about the message id, unless there is a user jump in to require that | 00:15 |
| Eva-i | hm, together with therve we decided it's not a small feature, it's a small bug | 00:16 |
| Eva-i | I guess it's half small feature and half bug | 00:16 |
| Eva-i | flwang: okay, maybe we can at least add queue name to notification? | 00:17 |
| flwang | Eva-i: yep, sure | 00:17 |
| *** malini has left #openstack-zaqar | 00:17 | |
| Eva-i | flwang: I'll modify the bug report then | 00:18 |
| flwang | Eva-i: cool, thanks :) | 00:18 |
| *** Qiming has quit IRC | 00:27 | |
| openstackgerrit | Fei Long Wang proposed openstack/zaqar: Fix the TTL issue of subscriptions for MongoDB https://review.openstack.org/270464 | 00:39 |
| flwang | Eva-i: as for the message id, it trigged me something | 01:03 |
| Eva-i | flwang: what it is? | 01:06 |
| flwang | kgriffs mentioned that we may should generate the message id by zaqar instead of by db | 01:06 |
| kgriffs | flwang actually, I saw somewhere the other day that they were having trouble synchronizing even atomic clocks with google spanner, but I think that if messages for any given queue are constrained to a cluster operating on the same local network, synchronizing to a local time server may give enough granularity. | 01:09 |
| kgriffs | would require further research and testing, but might be fruitful | 01:10 |
| flwang | kgriffs: do you think the python uuid lib can fit our requirement? | 01:10 |
| Eva-i | flwang: is so, in which Zaqar's driver layer the message id generation can be located? | 01:13 |
| flwang | Eva-i: depends if we want to pass it to notification driver | 01:14 |
| flwang | i mean the pipeline | 01:14 |
| *** Qiming has joined #openstack-zaqar | 01:22 | |
| *** flwang1 has quit IRC | 01:22 | |
| kgriffs | hmm, assuming you mean uuid version 1, that would give a theoretical granularity of 100 ns, and it will sort correctly | 01:25 |
| kgriffs | that granularity is more than enough | 01:26 |
| kgriffs | although clocks will be very very unlikely to be synced within a 100 ns | 01:26 |
| kgriffs | but if you could sync within 100 microseconds, that would support 10K id's created/sec without any collisions | 01:29 |
| kgriffs | (check my on my math) | 01:30 |
| kgriffs | to be honest, though, I wouldn't use the uuid module for this since it has a lot of checks and bit manipulation you don't need to waste CPU cycles on (not to mention it creates an object each time). Also, it takes up 128 bits in its binary form (more when encoded). | 01:33 |
| kgriffs | anyway, I gotta run, but there's some food for thought :) | 01:34 |
| *** flwang1 has joined #openstack-zaqar | 01:35 | |
| Eva-i | kgriffs: thank you | 01:36 |
| kgriffs | sure thing! | 01:36 |
| flwang | kgriffs: thank you | 01:38 |
| Eva-i | So I'm not sure, if I'm right, but if we want to include message id to notifications, we have two main options: 1. Generate message id in the transport layer 2. Do something with pipelines, for example, add additional pipeline after storage layer or make hook in storage layer, it can be done as decorator on storage layer methods. | 01:38 |
| Eva-i | Hm. Solution 1 is bad. Better not to give subscribers message ids, before adding messages to database. | 01:43 |
| flwang | Eva-i: so let's leave the message id for now and add a TODO in the code until we can figure out a good way to do that | 01:46 |
| Eva-i | flwang: want to hear a third solution I'm thinking of? If it's possible, it's the greatest solution. | 01:47 |
| Eva-i | *the third | 01:47 |
| flwang | ? | 01:48 |
| *** shakamunyi has joined #openstack-zaqar | 01:48 | |
| flwang | Eva-i: that's the 3rd? | 01:51 |
| Eva-i | flwang: not sure if it will work: 3. In the 'NotifierDriver' stage in a hook function make a manual call to the storage driver to get the result. Then inspect the result and take out message id. In the end of the hook function return the result, which will stop further moving through the pipeline (the same request will not get into the storage driver again) and allow transport layer to get the result. | 01:56 |
| Eva-i | flwang: of course in this case the user must ensure that the 'NotifierDriver' stage is the last in the pipeline, otherwise other stages, which the user has chosen in the pipeline, will not run. | 01:59 |
| flwang | it's the greatest? | 02:03 |
| Eva-i | flwang: it's the greatest solution a poor intern like me can imagine. | 02:04 |
| flwang | haha | 02:05 |
| flwang | Eva-i: we need a spec for that | 02:05 |
| flwang | we do have some tech debt, we can pay them one by one :) | 02:06 |
| Eva-i | flwang: so you like that? I can make a spec for that, but before I'll try to make a quick prototype. | 02:07 |
| Eva-i | to see if it works | 02:07 |
| flwang | sorry, i didn't mean i like that | 02:07 |
| flwang | it's complicated for me at the first glance | 02:07 |
| flwang | i mean we need more discussion | 02:07 |
| Eva-i | flwang: so you mean the discussion can be done during spec review? | 02:09 |
| *** shakamunyi has quit IRC | 02:10 | |
| *** shakamunyi has joined #openstack-zaqar | 02:10 | |
| *** shu-mutou has quit IRC | 02:15 | |
| flwang | Eva-i: or before drafting the spec | 02:15 |
| flwang | Eva-i: i don't have time now to dig into it, at the first glance, it's complicated than the bug described | 02:16 |
| flwang | so hence why i said we need more discussion to avoid wasting time | 02:16 |
| Eva-i | flwang: oki, It will be just few lines of code(<10). I'll test the prototype during weekend. | 02:18 |
| Eva-i | (on my free time) | 02:20 |
| flwang | cool | 02:22 |
| *** shu-mutou has joined #openstack-zaqar | 03:07 | |
| *** mdnadeem has joined #openstack-zaqar | 03:08 | |
| *** lei-zh has joined #openstack-zaqar | 03:08 | |
| Eva-i | flwang:if I'm not mistaken, Zaqar will drop support for redis in the future? | 03:38 |
| flwang | Eva-i: why you say that? | 03:41 |
| flaper87 | o.0 | 03:42 |
| flaper87 | no please | 03:42 |
| flaper87 | :D | 03:42 |
| Eva-i | flwang: I don't know, maybe I misheard someone telling someone about it. | 03:42 |
| flaper87 | Eva-i: we did that for sqlalchemy | 03:42 |
| flwang | flaper87: i can't believe you're around | 03:43 |
| Eva-i | flaper87: aha, I see, thank you | 03:43 |
| flaper87 | flwang: I'm always around | 03:43 |
| flaper87 | :D | 03:43 |
| flaper87 | Also, I'm in UTC-4:30 | 03:43 |
| flaper87 | i know, i know... it's hard to keep up with my traveling | 03:43 |
| flaper87 | :D | 03:43 |
| flaper87 | it's like 11:30pm now | 03:43 |
| flwang | not too late :) | 03:44 |
| Eva-i | for me it's 6:45 am =) | 03:45 |
| Eva-i | it's like I live in another country | 03:45 |
| Eva-i | going to sleep soon | 03:46 |
| Eva-i | is redis easy to install and start using it with Zaqar just like mongodb? I'd like to test some things. | 03:47 |
| flaper87 | Eva-i: it is | 03:48 |
| flaper87 | probably easier | 03:48 |
| Eva-i | flaper87: oki | 03:49 |
| flwang | flaper87: do you like the permanent subscription idea? | 03:51 |
| flaper87 | flwang: haven't read about it | 03:51 |
| flaper87 | where is it? | 03:51 |
| flaper87 | is it a spec? | 03:52 |
| flwang | if user don't pass in TTL for subscription, then it will be a permanent subscription | 03:52 |
| flwang | unless user delete it | 03:52 |
| flwang | https://review.openstack.org/#/c/270464/2 | 03:52 |
| flwang | flaper87: ^ | 03:52 |
| *** flwang1 has quit IRC | 04:01 | |
| *** lei-zh has quit IRC | 05:16 | |
| zigo | Hi there ! | 05:38 |
| zigo | There's no tag for zaqar-ui. | 05:38 |
| zigo | Could someone tag a Mitaka .0b2 release so I can package it in Debian? | 05:38 |
| zigo | flaper87: Can you do it? | 05:38 |
| zigo | Also, everything is called *-dashboard, it'd be nice to have zaqar-ui renamed as zaqar-dashboard too... | 05:39 |
| *** lei-zh has joined #openstack-zaqar | 05:46 | |
| *** boris-42 has joined #openstack-zaqar | 07:17 | |
| *** barra204 has joined #openstack-zaqar | 08:16 | |
| *** shakamunyi has quit IRC | 08:19 | |
| *** mdnadeem has quit IRC | 08:19 | |
| *** mdnadeem has joined #openstack-zaqar | 08:20 | |
| *** achanda has quit IRC | 09:01 | |
| *** Qiming has quit IRC | 09:54 | |
| *** rcernin has joined #openstack-zaqar | 10:18 | |
| *** lei-zh has quit IRC | 10:40 | |
| *** Qiming has joined #openstack-zaqar | 10:54 | |
| flaper87 | zigo: oh mmh, that's weird. I thin I can but it'll need the PTLs +1 anyway | 11:44 |
| flaper87 | think* | 11:44 |
| zigo | flaper87: There's no tag at all currently in that repo. | 12:31 |
| therve | zigo, It hasn't been released | 12:47 |
| therve | I don't think it's worth packaging just yet | 12:47 |
| zigo | therve: Could it be renamed zaqar-dashboard, just out of consistency? | 12:47 |
| therve | I don't know | 12:47 |
| zigo | Well, please consider it... :P | 12:48 |
| *** amitgandhinz has joined #openstack-zaqar | 13:54 | |
| *** pradk has quit IRC | 13:58 | |
| *** sriram has joined #openstack-zaqar | 14:11 | |
| *** pradk has joined #openstack-zaqar | 14:40 | |
| *** pradk has quit IRC | 14:47 | |
| *** pradk has joined #openstack-zaqar | 14:50 | |
| *** mdnadeem1 has joined #openstack-zaqar | 14:56 | |
| *** ametts has joined #openstack-zaqar | 15:06 | |
| *** malini has joined #openstack-zaqar | 15:23 | |
| *** Qiming has quit IRC | 15:24 | |
| *** achanda has joined #openstack-zaqar | 15:32 | |
| *** achanda has quit IRC | 15:43 | |
| *** amitgandhinz has quit IRC | 16:40 | |
| *** amitgandhinz has joined #openstack-zaqar | 16:41 | |
| *** mdnadeem1 has quit IRC | 16:58 | |
| *** mdnadeem1 has joined #openstack-zaqar | 17:02 | |
| *** csoukup has joined #openstack-zaqar | 17:08 | |
| *** mpanetta has joined #openstack-zaqar | 17:09 | |
| *** mdnadeem1 has quit IRC | 17:18 | |
| *** kgriffs is now known as kgriffs|afk | 17:21 | |
| *** kgriffs|afk is now known as kgriffs | 17:27 | |
| *** amitgandhinz has quit IRC | 17:52 | |
| *** amitgandhinz has joined #openstack-zaqar | 18:01 | |
| *** kgriffs is now known as kgriffs|afk | 18:06 | |
| *** boris-42 has quit IRC | 18:13 | |
| *** pradk has quit IRC | 18:20 | |
| *** csoukup has quit IRC | 18:33 | |
| *** pradk has joined #openstack-zaqar | 18:34 | |
| *** kgriffs|afk is now known as kgriffs | 18:36 | |
| *** AJaeger has joined #openstack-zaqar | 19:20 | |
| AJaeger | flwang: are you already awake? | 19:21 |
| *** mpanetta has quit IRC | 19:29 | |
| *** mpanetta has joined #openstack-zaqar | 19:30 | |
| *** malini has quit IRC | 20:42 | |
| AJaeger | seems we're not finding a common time, flwang ;( | 21:00 |
| *** rcernin has quit IRC | 21:02 | |
| Eva-i | AJaeger: Hi. Usually flwang is active this time. If he'll not appear tomorrow, you can probably consider writing an email to him. | 21:17 |
| Eva-i | AJaeger: hm... probably I'm wrong, because his weekend will start tomorrow | 21:18 |
| *** rcernin has joined #openstack-zaqar | 21:27 | |
| *** amitgandhinz has quit IRC | 21:43 | |
| *** flwang1 has joined #openstack-zaqar | 21:48 | |
| *** sriram has quit IRC | 21:58 | |
| *** Eva-i has quit IRC | 22:14 | |
| *** Eva-i has joined #openstack-zaqar | 22:42 | |
| *** flwang1 has quit IRC | 22:45 | |
| *** flwang1 has joined #openstack-zaqar | 22:55 | |
| *** Eva-i has quit IRC | 23:00 | |
| *** Eva-i has joined #openstack-zaqar | 23:00 | |
| *** flwang1 has quit IRC | 23:30 | |
| *** Qiming has joined #openstack-zaqar | 23:32 | |
| *** kgriffs is now known as kgriffs|afk | 23:44 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!