Changes between Version 156 and Version 157 of members/GRIP-UQAM


Ignore:
Timestamp:
Aug 4, 2016, 10:30:53 PM (8 years ago)
Author:
Grip Uqam
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • members/GRIP-UQAM

    v156 v157  
    262262http://stackoverflow.com/questions/3327312/drop-all-tables-in-postgresql
    263263 * http://stackoverflow.com/questions/20017552/delete-all-table-without-droping-database-in-postgres-in-django-dbshell-in-one-c
     264=== Postgres for cache tables ===
     265* nardberjean: dkg: may I ask you a question about postrgres and drupal cache tables?
     266* nardberjean: I read that posgres can be configured to be faster on less important data
     267* dkg: nardberjean: you can always ask questions here -- even though some of us might not know the answers ;)
     268* nardberjean: so I was asking me if that would apply to cache table
     269* dkg: yes, i recommend that cache tables be created as in-memory tables if you have enough RAM
     270* nardberjean: well I do not know it I can do that on mf machines
     271* nardberjean: as a member
     272* nardberjean: I have been told that the less RAM used the better on theses machines
     273* dkg: you might be interested in a few different points of documentation at a system level: [https://www.postgresql.org/docs/9.4/static/non-durability.html]
     274* dkg: in particular, look at [https://www.postgresql.org/docs/9.4/static/sql-createtable.html] and "unlogged"
     275* dkg: "the less RAM used the better" is always generally true but isn't very nuanced -- if you use less RAM by swapping everything to disk all the time to the point where no one can use the disk it seems worse to me.
     276* dkg: in my experience, most MF/PL machines are I/O constrained, not RAM constrained.
     277* nardberjean: well Non-Durable Settings would be good for I/O, and I could do it by myself, even if I feel the change should be made into Drupal code.
     278* nardberjean: for RAM tables, I do not know if it is already configured on mf machines, once I will have that info, I could try to move some small cache table to it...
     279* dkg: cool!  please report on support.mayfirst.org once you tried it out -- whether you make successful progress or otherwise, sorting this out is likely to be helpful to everyone on MF/PL. :)
     280* nardberjean: dkg: with d8 being very intense on cache, it makes it more an issue
     2812016-08-04
    264282== MongoDB ==
    265283* #11321