*** hoonetorg has quit IRC | 00:32 | |
*** gcb has quit IRC | 00:41 | |
*** hoonetorg has joined #openstack-oslo | 00:45 | |
*** zhangguoqing has joined #openstack-oslo | 00:48 | |
*** tovin07_ has joined #openstack-oslo | 00:59 | |
*** jamielennox is now known as jamielennox|away | 01:02 | |
*** pblaho has quit IRC | 01:11 | |
*** pblaho has joined #openstack-oslo | 01:12 | |
*** jamielennox|away is now known as jamielennox | 01:16 | |
*** e0ne has joined #openstack-oslo | 01:20 | |
*** david-lyle has joined #openstack-oslo | 01:24 | |
*** e0ne has quit IRC | 01:25 | |
*** hoonetorg has quit IRC | 01:42 | |
*** amotoki has joined #openstack-oslo | 01:46 | |
*** hoonetorg has joined #openstack-oslo | 01:54 | |
*** gcb has joined #openstack-oslo | 01:58 | |
*** amotoki has quit IRC | 01:58 | |
*** tommylikehu_ has joined #openstack-oslo | 02:03 | |
*** tommylikehu_ has quit IRC | 02:08 | |
*** yamahata has quit IRC | 02:14 | |
*** e0ne has joined #openstack-oslo | 02:21 | |
*** e0ne has quit IRC | 02:25 | |
*** yamamoto has joined #openstack-oslo | 02:29 | |
*** yamamoto has quit IRC | 02:33 | |
*** dharinic has joined #openstack-oslo | 02:34 | |
*** david-lyle has quit IRC | 02:36 | |
*** xianghui has joined #openstack-oslo | 02:40 | |
*** xianghui_ has joined #openstack-oslo | 02:43 | |
*** xianghui has quit IRC | 02:46 | |
*** ihrachys has joined #openstack-oslo | 02:46 | |
*** amotoki has joined #openstack-oslo | 03:00 | |
*** xianghui1__ has joined #openstack-oslo | 03:04 | |
*** xianghui1__ has quit IRC | 03:04 | |
*** xianghui_ has quit IRC | 03:07 | |
*** links has joined #openstack-oslo | 03:12 | |
*** nicolasbock has quit IRC | 03:15 | |
*** e0ne has joined #openstack-oslo | 03:22 | |
*** dimtruck is now known as zz_dimtruck | 03:23 | |
*** e0ne has quit IRC | 03:23 | |
*** erlon has quit IRC | 03:35 | |
*** david-lyle has joined #openstack-oslo | 03:37 | |
*** gcb has quit IRC | 03:45 | |
*** gcb has joined #openstack-oslo | 03:59 | |
*** Dinesh_Bhor has joined #openstack-oslo | 04:00 | |
*** zhangguoqing has quit IRC | 04:06 | |
*** e0ne has joined #openstack-oslo | 04:22 | |
*** e0ne has quit IRC | 04:27 | |
*** pksingh has joined #openstack-oslo | 05:08 | |
*** sanfern has joined #openstack-oslo | 05:10 | |
pksingh | Hello , i am getting 'oslo_config.cfg.NoSuchOptError: no such option verbose in group [DEFAULT]', in tox -epy27, yesterday it was working fine | 05:10 |
---|---|---|
pksingh | is there any changes related this in oslo? | 05:10 |
*** zhangguoqing has joined #openstack-oslo | 05:17 | |
*** e0ne has joined #openstack-oslo | 05:23 | |
*** e0ne has quit IRC | 05:28 | |
gcb | pksingh, which project? | 05:30 |
gcb | pksingh: we removed option verbose in https://review.openstack.org/#/c/444217/ | 05:31 |
pksingh | gcb: openstack/watcher | 05:32 |
pksingh | gcb: https://github.com/openstack/watcher/blob/master/watcher/tests/conf_fixture.py#L39, do we need to remove this option? | 05:32 |
*** zhangguoqing has quit IRC | 05:32 | |
*** zhangguoqing has joined #openstack-oslo | 05:34 | |
gcb | pksingh: yes, please remove the usage of option, then it should work | 05:34 |
*** yamamoto has joined #openstack-oslo | 05:35 | |
*** zhangguo_ has joined #openstack-oslo | 05:36 | |
*** zhangguoqing has quit IRC | 05:38 | |
pksingh | gcb: thanks a lot | 05:39 |
*** rcernin has joined #openstack-oslo | 05:39 | |
gcb | pksingh: np | 05:39 |
*** zhangguo_ has quit IRC | 05:39 | |
*** zhangguoqing has joined #openstack-oslo | 05:40 | |
*** zhangguoqing has quit IRC | 05:40 | |
*** yamamoto has quit IRC | 05:40 | |
*** zhangguoqing has joined #openstack-oslo | 05:41 | |
*** zhangguoqing has quit IRC | 05:45 | |
*** lpetrut has joined #openstack-oslo | 05:45 | |
*** yamahata has joined #openstack-oslo | 05:47 | |
*** zhangguoqing has joined #openstack-oslo | 05:48 | |
*** lpetrut has quit IRC | 05:49 | |
*** zhangguoqing has quit IRC | 05:49 | |
*** jamielennox is now known as jamielennox|away | 06:10 | |
*** zhangguoqing has joined #openstack-oslo | 06:11 | |
*** jaosorior has joined #openstack-oslo | 06:14 | |
*** nkrinner_afk is now known as nkrinner | 06:20 | |
*** pcaruana has joined #openstack-oslo | 06:24 | |
*** tovin07 has joined #openstack-oslo | 06:25 | |
*** jamielennox|away is now known as jamielennox | 06:25 | |
*** dharinic has quit IRC | 06:35 | |
*** yamamoto has joined #openstack-oslo | 06:41 | |
*** gcb has quit IRC | 06:42 | |
*** yamamoto has quit IRC | 06:47 | |
*** pksingh has quit IRC | 06:47 | |
*** oanson has quit IRC | 06:48 | |
*** jamielennox is now known as jamielennox|away | 06:50 | |
*** gcb has joined #openstack-oslo | 06:51 | |
*** jamielennox|away is now known as jamielennox | 06:57 | |
*** dharinic has joined #openstack-oslo | 07:01 | |
*** dharinic has quit IRC | 07:09 | |
*** tesseract has joined #openstack-oslo | 07:14 | |
openstackgerrit | Merged openstack/oslo.messaging master: Updated from global requirements https://review.openstack.org/442946 | 07:20 |
*** aarefiev_afk is now known as aarefiev | 07:32 | |
*** jpena|off is now known as jpena | 07:39 | |
*** yamamoto has joined #openstack-oslo | 07:43 | |
*** sileht has quit IRC | 07:45 | |
*** tovin07 has quit IRC | 07:45 | |
*** yamamoto has quit IRC | 07:48 | |
*** lucas-afk is now known as lucas-event | 07:49 | |
*** jpena is now known as jpena|off | 07:53 | |
*** e0ne has joined #openstack-oslo | 07:54 | |
openstackgerrit | Merged openstack/oslo.messaging master: Add HACKING.rst https://review.openstack.org/449391 | 07:55 |
*** jpena|off is now known as jpena | 07:59 | |
*** zzzeek has quit IRC | 08:00 | |
*** zhangguoqing has quit IRC | 08:00 | |
*** zzzeek has joined #openstack-oslo | 08:03 | |
*** shardy has joined #openstack-oslo | 08:03 | |
*** zhangguoqing has joined #openstack-oslo | 08:08 | |
*** tommylikehu_ has joined #openstack-oslo | 08:09 | |
*** tommylikehu_ has quit IRC | 08:13 | |
openstackgerrit | ChangBo Guo(gcb) proposed openstack/oslo.messaging master: Try to fix TestSerializer.test_call_serializer failed randomly https://review.openstack.org/451239 | 08:15 |
*** jamielennox is now known as jamielennox|away | 08:17 | |
openstackgerrit | ChangBo Guo(gcb) proposed openstack/oslo.messaging master: serializer: remove deprecated RequestContextSerializer https://review.openstack.org/447390 | 08:22 |
*** gcb has quit IRC | 08:39 | |
*** gcb has joined #openstack-oslo | 08:39 | |
*** yamamoto has joined #openstack-oslo | 08:45 | |
*** yamamoto has quit IRC | 08:50 | |
*** chenying__ has quit IRC | 08:57 | |
*** rpodolyaka_afk is now known as rpodolyaka | 08:57 | |
*** chenying__ has joined #openstack-oslo | 08:57 | |
*** ihrachys has quit IRC | 09:09 | |
*** ihrachys has joined #openstack-oslo | 09:10 | |
*** ihrachys has quit IRC | 09:17 | |
*** yamahata has quit IRC | 09:27 | |
*** sdague has joined #openstack-oslo | 09:29 | |
*** sambetts|afk is now known as sambetts | 09:30 | |
*** zhangguoqing has quit IRC | 09:32 | |
*** zhangguoqing has joined #openstack-oslo | 09:33 | |
*** amotoki has quit IRC | 09:40 | |
*** yamamoto has joined #openstack-oslo | 09:46 | |
*** zhangguoqing has quit IRC | 09:51 | |
*** yamamoto has quit IRC | 09:52 | |
*** zhangguoqing has joined #openstack-oslo | 09:53 | |
*** zhangguoqing has quit IRC | 09:55 | |
*** e0ne has quit IRC | 10:01 | |
*** e0ne has joined #openstack-oslo | 10:01 | |
*** tommylikehu_ has joined #openstack-oslo | 10:10 | |
*** tommylikehu_ has quit IRC | 10:15 | |
*** amotoki has joined #openstack-oslo | 10:16 | |
*** gcb has quit IRC | 10:16 | |
*** amotoki has quit IRC | 10:17 | |
*** tovin07_ has quit IRC | 10:20 | |
*** boden has joined #openstack-oslo | 10:28 | |
*** zhangguoqing has joined #openstack-oslo | 10:29 | |
*** nicolasbock has joined #openstack-oslo | 10:31 | |
*** amotoki has joined #openstack-oslo | 10:31 | |
openstackgerrit | Merged openstack/oslo.service master: Updated from global requirements https://review.openstack.org/451053 | 10:33 |
openstackgerrit | Merged openstack/oslo.service master: Fix some reST field lists in docstrings https://review.openstack.org/448484 | 10:34 |
*** amotoki has quit IRC | 10:40 | |
*** yamamoto has joined #openstack-oslo | 10:48 | |
openstackgerrit | Merged openstack/oslo.service master: Use Sphinx 1.5 warning-is-error https://review.openstack.org/448570 | 10:50 |
*** yamamoto has quit IRC | 10:54 | |
*** sanfern has quit IRC | 10:55 | |
*** yamamoto has joined #openstack-oslo | 11:09 | |
*** tommylikehu_ has joined #openstack-oslo | 11:10 | |
*** amotoki has joined #openstack-oslo | 11:11 | |
*** tommylikehu_ has quit IRC | 11:18 | |
*** yamamoto has quit IRC | 11:19 | |
*** tommylikehu_ has joined #openstack-oslo | 11:22 | |
*** Guest16838 has quit IRC | 11:26 | |
*** Guest16838 has joined #openstack-oslo | 11:29 | |
*** tommylikehu_ has quit IRC | 11:30 | |
openstackgerrit | Merged openstack/tooz master: tests: fix etcd and consul test run https://review.openstack.org/448564 | 11:37 |
*** sanfern has joined #openstack-oslo | 11:47 | |
*** tommylikehu_ has joined #openstack-oslo | 11:56 | |
*** tommylikehu_ has quit IRC | 11:57 | |
*** zhangguoqing has quit IRC | 11:59 | |
*** zhangguoqing has joined #openstack-oslo | 11:59 | |
*** zhangguoqing has quit IRC | 12:03 | |
openstackgerrit | Merged openstack/oslo.messaging master: Trivial: Add executor 'threading' in docstring https://review.openstack.org/447424 | 12:18 |
*** yamamoto has joined #openstack-oslo | 12:20 | |
*** bjornar_ has joined #openstack-oslo | 12:20 | |
*** gordc has joined #openstack-oslo | 12:23 | |
*** yamamoto has quit IRC | 12:25 | |
*** tommylikehu_ has joined #openstack-oslo | 12:28 | |
openstackgerrit | Dolph Mathews proposed openstack/oslo.config master: Unit tests to illustrate positional argument bug https://review.openstack.org/450956 | 12:36 |
openstackgerrit | Dolph Mathews proposed openstack/oslo.config master: Support required positional arguments correctly https://review.openstack.org/450928 | 12:36 |
*** jpena is now known as jpena|lunch | 12:40 | |
*** ansmith has joined #openstack-oslo | 12:46 | |
*** dtardivel has joined #openstack-oslo | 13:01 | |
*** gordc has quit IRC | 13:02 | |
*** gordc has joined #openstack-oslo | 13:02 | |
*** lucas-event is now known as lucasagomes | 13:03 | |
*** ihrachys has joined #openstack-oslo | 13:05 | |
*** kgiusti has joined #openstack-oslo | 13:05 | |
*** ihrachys has quit IRC | 13:10 | |
*** sileht has joined #openstack-oslo | 13:19 | |
*** links has quit IRC | 13:20 | |
*** erlon has joined #openstack-oslo | 13:20 | |
*** yamamoto has joined #openstack-oslo | 13:22 | |
rpodolyaka | bjornar_: hey! I noticed that you pinged me about handling of multi-server DB configurations in oslo.db on Monday while I was away | 13:25 |
rpodolyaka | bjornar_: I know that zzzeek has been working on something called haalchemy which must be what you want to see | 13:25 |
rpodolyaka | bjornar_: take a look at https://bitbucket.org/zzzeek/haalchemy | 13:26 |
*** yamamoto has quit IRC | 13:28 | |
rpodolyaka | bjornar_: I'm not sure how mature it is, I just took a quick a look at the docs today, but I like the idea that multi-server config is going to be transparent to clients like oslo.db | 13:28 |
bjornar_ | rpodolyaka, looks good, but is the project still aliive? | 13:29 |
rpodolyaka | bjornar_: not sure :( | 13:30 |
bjornar_ | rpodolyaka, what we do currently is to have loadbalancers-everywhere, but it should not really be needed, but better than the central lb approach | 13:30 |
*** kiseok7_ has joined #openstack-oslo | 13:32 | |
*** kiseok7_ has left #openstack-oslo | 13:33 | |
*** tommylikehu_ has quit IRC | 13:34 | |
*** gcb has joined #openstack-oslo | 13:36 | |
*** jaosorior has quit IRC | 13:38 | |
*** dmsimard is now known as dmsimard|afk | 13:39 | |
*** e0ne has quit IRC | 13:41 | |
*** e0ne has joined #openstack-oslo | 13:41 | |
*** pkoniszewski has left #openstack-oslo | 13:42 | |
*** jaosorior has joined #openstack-oslo | 13:43 | |
*** jpena|lunch is now known as jpena | 13:45 | |
*** tommylikehu_ has joined #openstack-oslo | 13:47 | |
*** tongli has joined #openstack-oslo | 13:51 | |
*** e0ne has quit IRC | 13:54 | |
openstackgerrit | Merged openstack-dev/pbr master: Lazy import pkg_resources https://review.openstack.org/450913 | 13:54 |
*** lpetrut has joined #openstack-oslo | 13:55 | |
*** eck` is now known as eck`gone | 14:03 | |
*** eck`gone is now known as eck` | 14:03 | |
*** nicolasbock has quit IRC | 14:05 | |
*** tommylikehu_ has quit IRC | 14:09 | |
*** tommylikehu_ has joined #openstack-oslo | 14:10 | |
*** yamamoto has joined #openstack-oslo | 14:24 | |
*** yamamoto has quit IRC | 14:29 | |
*** ihrachys has joined #openstack-oslo | 14:30 | |
*** zz_dimtruck is now known as dimtruck | 14:30 | |
zzzeek | rpodolyaka: I'm likely not continuing w/ haalchemy for now, check out http://www.proxysql.com/ | 14:42 |
rpodolyaka | bjornar_: ^ | 14:43 |
sileht | / | 14:44 |
rpodolyaka | although, I guess what bjornar_ is looking for is no proxy in the middle, but rather just the connection string pointing to multiple servers | 14:45 |
bjornar_ | zzzeek, bah | 14:45 |
zzzeek | bjornar_: why | 14:45 |
bjornar_ | zzzeek, this is not the right solution, the application should be able to handle upstream failure anyway | 14:45 |
bjornar_ | zzzeek, a proxy is a unnessesary bottleck, requires extra hops, extra latency, extra configuration/scaleout complexity and is a spof unless multiple ips are used (which it cannot) | 14:47 |
zzzeek | bjornar_: the most recent haalchemy concept included a service that runs to report health on the DB, it just doesn't go "between" the application and database. but was still required. an earlier concept had each application process pinging the database independently. that approach has a lot of disadvantages | 14:47 |
zzzeek | bjornar_: are we talking about openstack w/ galera ? | 14:47 |
bjornar_ | zzzeek, for example, but could be anything basically with multiple ip's .. most relevant is perhaps galera, but not directly related | 14:48 |
zzzeek | bjornar_: and you have load-profiled haproxy and found it to be a bottleneck ? | 14:48 |
bjornar_ | we are basically talking about a connection pool with multiple servers/ips -- where it points to is irelevant -> could even point to the haproxy proxy for all I care | 14:48 |
zzzeek | bjornar_: b.c. that's a huge claim. the latency of haproxy etc. is incredibly low | 14:48 |
bjornar_ | zzzeek, its irelevant. Its extra hops, extra complexity, extra software, spof and so on | 14:49 |
bjornar_ | and if the implementation of a connection pool is beyond scope of oslo.db developers, I hope you are actively trying to find new ones. | 14:50 |
zzzeek | bjornar_: building a database pinger into every openstack process is more complex and more software (E.g. it is custom to openstack and needs a maintainer) | 14:50 |
bjornar_ | its not about every-openstack project, its about oslo.db or eventually sqlalchemy or whatever dependency | 14:50 |
bjornar_ | memcached and rabbit already has this | 14:50 |
rpodolyaka | bjornar_: curiously enough, it's one of the pain points with the memcache | 14:51 |
zzzeek | bjornar_: not clear on what "this" is at this point. HAAlchemy had lots and lots of things it wanted to do, not sure which specific things you are looking for. do you ahve a document ? | 14:51 |
rpodolyaka | when one (or more) of the memcached nodes go down we wait $read_timeout seconds on each request to it | 14:52 |
rpodolyaka | until we mark it as "dead" | 14:52 |
zzzeek | bjornar_: but I will say, "I hope you as maintainer of project X are looking for more people to help with your project because you don't have enough features" is not a helpful statement | 14:52 |
rpodolyaka | and that knowledge is not shared between forks of openstack services - each one must keep its own list of "dead" memcached's | 14:53 |
bjornar_ | rpodolyaka, sure, the algo can be tuned, but same goes for a keepalived whatever failover for a virtual ip or how one does it | 14:53 |
zzzeek | bjornar_: direct application->db w/ in-app failover is interesting but is very hard to justify given the existing approaches. people dont really care about "hops" unless you can show measurable degradation in performance | 14:54 |
bjornar_ | zzzeek, its not just about latency and hops, its also about failover scenarios and so on. Complexity of the setup as a whole | 14:54 |
zzzeek | bjornar_: that said, haalchemy was going for that approahc where apps connect to the DB directly, the service is used for coordiantion. but it is *very* complicated | 14:55 |
bjornar_ | say one wants to use two loadbalancers for example, its not currently possible | 14:55 |
zzzeek | bjornar_: you'd use haproxy | 14:55 |
bjornar_ | zzzeek, so how do I ask for example keystone to connect to two haproxy instances? | 14:56 |
zzzeek | bjornar_: put another one in between | 14:56 |
*** amotoki has quit IRC | 14:56 | |
bjornar_ | zzzeek, its not that complex, and most of the logic is already inside sqlalchemy.. I mean, it does the pings and reconnects, the whole difference is _where_ it reconnects | 14:57 |
bjornar_ | zzzeek, stil a single one, see | 14:57 |
rpodolyaka | service discovery via dns? like consul? | 14:57 |
bjornar_ | so when haproxy fails for some reason, you move it to a different server, and sqlalchemy is able to reconnect | 14:57 |
zzzeek | bjornar_: theres a lot of things we need in this area and the current pinging it does is not really sufficient. | 14:58 |
bjornar_ | most of this can just be done on the tcp layer | 14:58 |
zzzeek | bjornar_: the pinging needs to be asynchronous in a background thread | 14:58 |
bjornar_ | does not even need to know protocol | 14:58 |
zzzeek | bjornar_: also, connection pooling among hundreds of independent processes is a big use case, i wanted to tackle that also | 14:58 |
zzzeek | haalchemy was going to try some very ambitious way of doing this but proxysql makes this easy, working similarly to pgbouncer | 14:59 |
bjornar_ | glb is probably the better option | 14:59 |
bjornar_ | but should not be needed anyway I think | 14:59 |
bjornar_ | currently we spawn pen on every container, and exposes a socket for sqlalchemy to connect to | 15:00 |
zzzeek | bjornar_: please read https://bitbucket.org/zzzeek/haalchemy/src/6623f93c122f5e4ac9de208c09512699e30a76f8/docs/build/index.rst?at=master&fileviewer=file-view-default to see what i was working on. it's hard to justify given the existence of proxysql - if the only difference is "hops" | 15:00 |
bjornar_ | zzzeek, the connection pooling in the application side should just be simple, I mean.. either a host listens on tcp port X so one can establish a connection, or it does not | 15:01 |
bjornar_ | zzzeek, this is needed even if haproxy | 15:01 |
openstackgerrit | Julien Danjou proposed openstack/tooz master: redis: fix concurrent access on acquire() https://review.openstack.org/447199 | 15:01 |
bjornar_ | then leave the complexity of figuring out if the application layer is really alive to the "proxy", and make it just not reject connections or not advertise it's ip if its not feeling well | 15:02 |
*** pblaho has quit IRC | 15:02 | |
bjornar_ | zzzeek, in that way, it does not really have to work differently (as I said above) to how sqlalchemy currently treats a haproxy or whatever that goes down and comes up again --> only difference is that it should be able to connect to multiple ips (I dont think it even does that with dns and multiple A-records) | 15:04 |
bjornar_ | zzzeek, this document you provided is far too complex, and as complex projects go, they seldom get finished, and contain lots of bugs if they do | 15:04 |
bjornar_ | zzzeek, the main thing "we" need is beeing able to spesify multiple ip's, and have the db-layer reconnect to the next server (and some kind of hashing) if the connection is "dead" | 15:05 |
*** rcernin has quit IRC | 15:05 | |
zzzeek | bjornar_: the haalchemy initial versions did that on top of plain vanilla sqlalchemy, you can use a custom "connect" routine to get this behavior right now | 15:08 |
zzzeek | bjornar_: the problem w/ that kind of routing for "us" is that, if a DB has been restarted, the pool has a stale connection. it looks "dead". but, we dont want to go to the next node in that case, we want to first test the *same* node to see if its back up | 15:10 |
zzzeek | bjornar_: all of this connection testing, if you use the current "ping" in oslo.db, happens up front when someone is trying to actually make a request to do something | 15:10 |
zzzeek | bjornar_: for "us" also, it's important that all applications connect to the *same* database at once. we need coordination to make sure they are all doing the same thing | 15:11 |
*** dmsimard|afk is now known as dmsimard | 15:11 | |
zzzeek | if you dont need that, then you can provide a connect() routine that just connects to whichever host is up, that's a simple extension | 15:12 |
*** e0ne has joined #openstack-oslo | 15:13 | |
bjornar_ | huh.. if you need that, you just spesify a single server, right? | 15:13 |
bjornar_ | I dont see the relevance | 15:13 |
bjornar_ | zzzeek, yeah, so thats what oslo.db should implement, I think.. connecting to multple host, a timeout value for "dead" hosts and a hashing based on whatever | 15:16 |
*** mtanino has joined #openstack-oslo | 15:16 | |
bjornar_ | ?intial_server=1234 | 15:16 |
bjornar_ | and then server = initial_server % num_servers | 15:17 |
zzzeek | bjornar_: this is like a "micro" load balancer feature that serves a very narrow set of use cases. feel free to propose a gerrit as an oslo.db feature | 15:18 |
bjornar_ | I dont think reconnecting on failure and exposing multiple server is a narrow use case, as it just opens possibilities that are more advanced | 15:19 |
zzzeek | bjornar_: as haalchemy does, you intercept the engine's creation, the most modern way is using http://docs.sqlalchemy.org/en/latest/core/connections.html?highlight=createengineplugin#sqlalchemy.engine.CreateEnginePlugin | 15:19 |
zzzeek | bjornar_: it's narrow because it doesnt allow for coordination among processes. it only allows "everyone connects to many hosts" balancing. it doesn't provide for, "Everyone connects to host A, then host B, atomically" | 15:20 |
zzzeek | bjornar_: a proxy server OTOH can ensure everyone does the same thing | 15:20 |
bjornar_ | zzzeek, who would coordinate along processes and why? | 15:20 |
*** pblaho has joined #openstack-oslo | 15:20 | |
zzzeek | bjornar_: in our case it's because we keep all of openstack on one galera master at a time to avoid failed transactions | 15:21 |
zzzeek | bjornar_: as well as replication lag | 15:21 |
bjornar_ | zzzeek, this can still be done with multiple ip's .. you dont need to change your proxy configuration | 15:21 |
bjornar_ | zzzeek, but: you can have multiple proxies doing the same thing, and faster failover, better scalability and ha | 15:22 |
*** yamahata has joined #openstack-oslo | 15:22 | |
bjornar_ | and lets face it.. database sync is not the scalability/latency issue in openstack -- its the braindead developers writing sql ;) | 15:23 |
zzzeek | bjornar_: i dont understand what "multiple ips / multiple proxies" refers to | 15:23 |
bjornar_ | so, currently you probably have a single haproxy instance that you send your sql queries to.. | 15:24 |
zzzeek | bjornar_: galera has replication lag. it can cause failures right now | 15:24 |
bjornar_ | and from here you do some black-magic to keep all the connection on one galera backend | 15:24 |
bjornar_ | you can still do the same with multiple proxies -- the way your loadbalancing/proxy server is configured/synchronized is beyond the scope of having the ability to connect to multiple ip's | 15:25 |
bjornar_ | zzzeek, where is the documentation on the bug that can cause failures? | 15:25 |
zzzeek | bjornar_: the ability to connect to multiple ips is unnecessary if you just use proxies | 15:26 |
*** yamamoto has joined #openstack-oslo | 15:26 | |
zzzeek | bjornar_: https://www.percona.com/blog/2013/03/03/investigating-replication-latency-in-percona-xtradb-cluster/ | 15:26 |
bjornar_ | zzzeek, why do you say that? | 15:26 |
bjornar_ | 2013 | 15:26 |
zzzeek | bjornar_: we put the multiple database hosts behind the proxy | 15:26 |
*** tongli has quit IRC | 15:26 | |
bjornar_ | zzzeek, sure, but why not use multiple ip's? you see the fact that loadbalancer scale and ha can be improved this way without affecting your l7 balancing | 15:28 |
bjornar_ | zzzeek, ask yourself: where is your spof | 15:28 |
zzzeek | bjornar_: does "multiple ips" mean, "embed multiple ips in the client" like you are proposing? | 15:28 |
bjornar_ | sure | 15:29 |
bjornar_ | the client is the only place where you can do this in a scalable way without a spof | 15:29 |
zzzeek | bjornar_: it is difficult to ensure that all python processes use the *same* host at exactly the same time, especially when a node fails | 15:29 |
bjornar_ | the client _needs_ to be able to use multiple ip's .. loadbalancer or not | 15:29 |
zzzeek | bjornar_: i had it doing this using a fixed scheme | 15:29 |
zzzeek | bjornar_: but it still doesnt work unless you have an asynchronous pinger | 15:30 |
bjornar_ | zzzeek, it it is that dangerous for you, you should have only a master/slave setup -- that stuff you say there is just plain BS | 15:30 |
*** yamamoto has quit IRC | 15:31 | |
bjornar_ | zzzeek, so how does sqlalchemy work on proxy down today? | 15:31 |
bjornar_ | zzzeek, will it be any difference with multiple ip's? | 15:31 |
zzzeek | bjornar_: problem. process A is very busy. database node 1 is stopped and restarted. process A's pinger has stale pooled connection, it pings, sees DB as down (But its actually up again). LB routine bounces it to database #2. slower Python process B decides to do some work, connects to DB 1, sees nothing wrong. the scheme has now failed | 15:31 |
bjornar_ | zzzeek, this is your very spesific setup | 15:32 |
zzzeek | bjornar_: from my POV *you* have a very specific setup :) | 15:32 |
bjornar_ | zzzeek, and as I said: you do not need to connect to the database directly even if you can spesify multiple ip's | 15:32 |
zzzeek | where you can go to any node at any time | 15:32 |
bjornar_ | zzzeek, the multiple ips ca all reside on a single haproxy for all I care, or synchronized haproxies that all select the same active server, or whatever you want --- out of scope for the simple fact of spesifying multiple upstream servers | 15:33 |
*** eliqiao has quit IRC | 15:33 | |
bjornar_ | zzzeek, so having multiple A-records for a domain is very special? | 15:33 |
zzzeek | bjornar_: anyway, what you want is super easy to build. look into CreateEnginePlugin and then set up a pool creator() function and you're good to go | 15:33 |
zzzeek | bjornar_: for an internal application database connection URL, yes that's fairly uncommon | 15:34 |
zzzeek | bjornar_: but also, easy to support | 15:36 |
*** eliqiao has joined #openstack-oslo | 15:37 | |
bjornar_ | its not about the problem of building it, its the fact that it should be included in oslo.db | 15:38 |
openstackgerrit | Merged openstack/oslo.config master: Add HACKING.rst https://review.openstack.org/449409 | 15:41 |
zzzeek | bjornar_: propose a gerrit to oslo.db. we have that nutty NDB thing going in I am bothered a lot more by that :) | 15:42 |
zzzeek | bjornar_: this only bothers me b.c. i can see people thinking it does a lot more than it does | 15:42 |
*** lpetrut has quit IRC | 15:42 | |
openstackgerrit | ChangBo Guo(gcb) proposed openstack/oslo.config master: Make method set_override's parameter enforce_type=True by default https://review.openstack.org/328692 | 15:43 |
*** chenying__ has quit IRC | 15:43 | |
zzzeek | bjornar_: openstack is, "open". if your issue is, "openstack should have feature X and *someone in openstack needs to write it for us*", that's not as much the culture :) | 15:43 |
*** chenying__ has joined #openstack-oslo | 15:44 | |
zzzeek | bjornar_: i will gladly +2 a well written and tested round-robin url plugin for oslo.db if that's all we're talking about | 15:44 |
*** eliqiao has quit IRC | 15:46 | |
*** nkrinner is now known as nkrinner_afk | 15:51 | |
bjornar_ | zzzeek, sure, its just about doing some basic clientside ha, what one uses it for is beyond scope | 15:53 |
*** eliqiao has joined #openstack-oslo | 15:53 | |
bjornar_ | zzzeek, I think it's more about getting it up for discussion. I think the reasons for having it outweighs the reasons for _not_ having it. | 15:54 |
bjornar_ | And actually I was surprised that not sqlalchemy nor oslo.db provided this func already | 15:55 |
*** tommylikehu_ has quit IRC | 15:56 | |
*** tommylikehu_ has joined #openstack-oslo | 15:57 | |
*** amotoki has joined #openstack-oslo | 15:57 | |
*** tesseract has quit IRC | 15:58 | |
*** bjornar_ has quit IRC | 16:01 | |
openstackgerrit | John Perkins proposed openstack/oslo.config master: Oslo Validator https://review.openstack.org/384559 | 16:01 |
*** amotoki has quit IRC | 16:03 | |
*** e0ne has quit IRC | 16:06 | |
*** e0ne has joined #openstack-oslo | 16:06 | |
*** aarefiev is now known as aarefiev_afk | 16:08 | |
*** dmsimard is now known as dmsimard|afk | 16:13 | |
*** d0ugal has quit IRC | 16:17 | |
*** Rockyg has joined #openstack-oslo | 16:24 | |
*** tommylikehu_ has quit IRC | 16:25 | |
*** yamamoto has joined #openstack-oslo | 16:27 | |
*** yamamoto has quit IRC | 16:33 | |
*** shardy has quit IRC | 16:35 | |
*** e0ne has quit IRC | 16:38 | |
*** lpetrut has joined #openstack-oslo | 16:38 | |
*** d0ugal has joined #openstack-oslo | 16:38 | |
*** d0ugal has quit IRC | 16:43 | |
*** kzaitsev_ws has quit IRC | 16:44 | |
*** lucasagomes is now known as lucas-afk | 16:49 | |
*** pblaho has quit IRC | 16:50 | |
*** yamahata has quit IRC | 16:50 | |
*** shardy has joined #openstack-oslo | 16:52 | |
*** amotoki has joined #openstack-oslo | 16:59 | |
*** eliqiao has quit IRC | 17:01 | |
*** eliqiao has joined #openstack-oslo | 17:02 | |
openstackgerrit | Merged openstack/oslo.messaging master: [zmq] pass a dummy TransportURL to register_opts https://review.openstack.org/449172 | 17:03 |
*** amotoki has quit IRC | 17:04 | |
*** pblaho has joined #openstack-oslo | 17:07 | |
*** links has joined #openstack-oslo | 17:07 | |
*** lpetrut has quit IRC | 17:07 | |
*** sambetts is now known as sambetts|afk | 17:11 | |
*** yamahata has joined #openstack-oslo | 17:13 | |
*** links has quit IRC | 17:19 | |
*** jpena is now known as jpena|off | 17:20 | |
*** lpetrut has joined #openstack-oslo | 17:21 | |
openstackgerrit | Merged openstack/oslo.messaging master: Fix simulator's use of Notifier - use 'topics' not 'topic' https://review.openstack.org/447559 | 17:22 |
*** yamamoto has joined #openstack-oslo | 17:29 | |
*** bjornar_ has joined #openstack-oslo | 17:32 | |
*** yamamoto has quit IRC | 17:35 | |
openstackgerrit | John Perkins proposed openstack/oslo.config master: Oslo Validator https://review.openstack.org/384559 | 17:37 |
*** mtanino has quit IRC | 17:40 | |
*** e0ne has joined #openstack-oslo | 17:42 | |
openstackgerrit | Merged openstack/tooz master: tests: tests fail if no URL is set + run partitioner tests on basic drivers https://review.openstack.org/448568 | 17:44 |
*** eliqiao has quit IRC | 17:49 | |
*** eliqiao has joined #openstack-oslo | 17:50 | |
*** pcaruana has quit IRC | 17:52 | |
*** e0ne has quit IRC | 17:53 | |
*** lpetrut has quit IRC | 18:00 | |
*** dtardivel has quit IRC | 18:00 | |
*** amotoki has joined #openstack-oslo | 18:00 | |
*** amotoki has quit IRC | 18:05 | |
*** bjornar_ has quit IRC | 18:08 | |
*** mtanino has joined #openstack-oslo | 18:09 | |
openstackgerrit | Merged openstack/tooz master: redis: fix concurrent access on acquire() https://review.openstack.org/447199 | 18:12 |
*** yamamoto has joined #openstack-oslo | 18:31 | |
*** yamamoto has quit IRC | 18:36 | |
*** openstackstatus has joined #openstack-oslo | 18:45 | |
*** ChanServ sets mode: +v openstackstatus | 18:45 | |
*** e0ne has joined #openstack-oslo | 18:50 | |
*** gordc has quit IRC | 18:59 | |
*** amotoki has joined #openstack-oslo | 19:01 | |
openstackgerrit | Sean Dague proposed openstack/oslo.log master: WIP: this is all the hackiest hackness of systemd integration https://review.openstack.org/451525 | 19:02 |
*** amotoki has quit IRC | 19:05 | |
*** e0ne has quit IRC | 19:06 | |
*** Guest16838 has quit IRC | 19:14 | |
openstackgerrit | Stephen Finucane proposed openstack/oslo.config master: generator: Support outputting in JSON format https://review.openstack.org/451529 | 19:15 |
*** david-lyle has quit IRC | 19:19 | |
*** david-lyle has joined #openstack-oslo | 19:19 | |
*** Guest16838 has joined #openstack-oslo | 19:25 | |
*** bjornar_ has joined #openstack-oslo | 19:26 | |
*** david-lyle has quit IRC | 19:27 | |
*** eliqiao has quit IRC | 19:28 | |
*** david-lyle has joined #openstack-oslo | 19:30 | |
*** yamamoto has joined #openstack-oslo | 19:32 | |
*** openstackgerrit has quit IRC | 19:33 | |
*** jamielennox|away is now known as jamielennox | 19:33 | |
*** eliqiao has joined #openstack-oslo | 19:33 | |
*** yamamoto has quit IRC | 19:38 | |
*** syjulian has quit IRC | 19:44 | |
*** d0ugal has joined #openstack-oslo | 19:49 | |
*** d0ugal has quit IRC | 20:00 | |
*** amotoki has joined #openstack-oslo | 20:02 | |
*** amotoki has quit IRC | 20:06 | |
*** dimtruck is now known as zz_dimtruck | 20:21 | |
*** openstackgerrit has joined #openstack-oslo | 20:26 | |
openstackgerrit | Sean Dague proposed openstack/oslo.log master: WIP: this is all the hackiest hackness of systemd integration https://review.openstack.org/451525 | 20:26 |
*** Guest6666 has quit IRC | 20:28 | |
*** Guest6666 has joined #openstack-oslo | 20:29 | |
*** ansmith has quit IRC | 20:32 | |
*** david-lyle has quit IRC | 20:33 | |
*** yamamoto has joined #openstack-oslo | 20:34 | |
*** david-lyle has joined #openstack-oslo | 20:35 | |
*** yamamoto has quit IRC | 20:40 | |
*** david-lyle has quit IRC | 20:40 | |
*** zz_dimtruck is now known as dimtruck | 20:42 | |
*** david-lyle has joined #openstack-oslo | 20:42 | |
*** david-lyle has quit IRC | 20:46 | |
*** david-lyle has joined #openstack-oslo | 20:52 | |
*** syjulian has joined #openstack-oslo | 20:58 | |
*** amotoki has joined #openstack-oslo | 21:02 | |
*** amotoki has quit IRC | 21:07 | |
*** kgiusti has left #openstack-oslo | 21:15 | |
*** ihrachys has quit IRC | 21:26 | |
*** ansmith has joined #openstack-oslo | 21:33 | |
*** yamamoto has joined #openstack-oslo | 21:36 | |
*** bjornar_ has quit IRC | 21:36 | |
openstackgerrit | John Perkins proposed openstack/oslo.config master: Oslo Validator https://review.openstack.org/384559 | 21:37 |
*** dimtruck is now known as zz_dimtruck | 21:41 | |
*** yamamoto has quit IRC | 21:42 | |
*** dmitryme has quit IRC | 21:43 | |
*** DinaBelova has quit IRC | 21:44 | |
*** Guest16838 has quit IRC | 21:45 | |
*** dmitryme has joined #openstack-oslo | 21:46 | |
*** DinaBelova has joined #openstack-oslo | 21:46 | |
*** Guest16838 has joined #openstack-oslo | 21:47 | |
*** shardy has quit IRC | 21:50 | |
*** erlon has quit IRC | 21:55 | |
*** boden has quit IRC | 22:00 | |
*** amotoki has joined #openstack-oslo | 22:03 | |
*** bjornar_ has joined #openstack-oslo | 22:06 | |
*** amotoki has quit IRC | 22:08 | |
*** jamielennox is now known as jamielennox|away | 22:09 | |
*** jamielennox|away is now known as jamielennox | 22:12 | |
*** zz_dimtruck is now known as dimtruck | 22:17 | |
*** yamamoto has joined #openstack-oslo | 22:38 | |
*** bjornar_ has quit IRC | 22:42 | |
*** yamamoto has quit IRC | 22:42 | |
*** sdague has quit IRC | 22:46 | |
*** tommylikehu_ has joined #openstack-oslo | 22:51 | |
*** tommylikehu_ has quit IRC | 22:56 | |
*** amotoki has joined #openstack-oslo | 23:04 | |
*** tommylikehu_ has joined #openstack-oslo | 23:06 | |
*** kornica has joined #openstack-oslo | 23:06 | |
*** kornicameister has quit IRC | 23:08 | |
*** amotoki has quit IRC | 23:09 | |
*** tommylikehu_ has quit IRC | 23:10 | |
*** yamamoto has joined #openstack-oslo | 23:39 | |
*** yamamoto has quit IRC | 23:44 | |
*** hieulq has quit IRC | 23:47 | |
openstackgerrit | Merged openstack/oslo.config master: Use Sphinx 1.5 warning-is-error https://review.openstack.org/446589 | 23:49 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!