Closed domoritz closed 10 years ago
Check out jsonQueries/scalability_valmeida/create_deployment.py
. I think that's where Victor put his script to create a Postgres-enabled deployment file.
Here's a snip from the production Myria config that does this:
% grep aldebaran myria-production.cfg.psql
1 = aldebaran.cs.washington.edu:1778:/disk1/myria-production:myria1
19 = aldebaran.cs.washington.edu:1779:/disk2/myria-production:myria2
37 = aldebaran.cs.washington.edu:1780:/disk3/myria-production:myria3
55 = aldebaran.cs.washington.edu:1781:/disk4/myria-production:myria4
May be worth moving that .py
file to myriadeploy
and possibly reorganizing myriadeploy
itself. I think it's a hodgepodge of scripts made by and modified by several different authors.
Agreed. A single script that delegates work and does some sanity checks might be useful as well.
Hardcoding the postgres port to 5401 doesn't seem to be a good idea anyway.
I found this comment:
// TODO: Allow using the parameters to create the connection info.
// Now it is hardcoded to use a specific connection info, which allows only one
// myria instance per machine in the cluster
Sounds like it is not even possible to have multiple workers on one machine with pg (and probably mysql or monet as well).
There are four instances on aldebaran per the snip I sent you. They just have to use different databases. This is a single pg instance, just different databases.
On Mon, Feb 3, 2014 at 1:15 PM, Dominik Moritz notifications@github.comwrote:
Hardcoding the postgres port to 5401 doesn't seem to be a good idea anyway.
I found this comment:
// TODO: Allow using the parameters to create the connection info. // Now it is hardcoded to use a specific connection info, which allows only one // myria instance per machine in the cluster
Sounds like it is not even possible to have multiple workers on one machine with pg (and probably mysql or monet as well).
Reply to this email directly or view it on GitHubhttps://github.com/uwescience/myria/issues/393#issuecomment-34000771 .
For reference
WORKER_NUMBER = SERVER:PORT:DIRECTORY:DATABASE
(closable?)
A deployment with
Does not work because when we insert into the same database, it causes conflicts.
Is there a way to define a different connection per worker in the config file?