| Time |
Nick |
Message |
| 08:43 |
|
mmorgan joined #evergreen |
| 08:47 |
|
dguarrac joined #evergreen |
| 09:13 |
|
mmorgan1 joined #evergreen |
| 09:18 |
|
mmorgan joined #evergreen |
| 09:18 |
|
mmorgan left #evergreen |
| 09:55 |
|
mmorgan joined #evergreen |
| 10:14 |
csharp_ |
anybody out there experimenting with partitioning larger tables in EG databases? I'm trying to get seq-scan-style lookups to move faster and indexes might not be the best option |
| 10:14 |
* csharp_ |
glances meaningfully at eeevil |
| 10:15 |
berick |
curious which table(s) |
| 10:16 |
eeevil |
csharp_: depends on the specific goals. for reporting? yeah, go for it, especially for the aged circs and holds tables |
| 10:16 |
csharp_ |
berick: right now I'm looking at our "delete dormant patron accounts" process that we are hoping to run nightly |
| 10:17 |
berick |
ah, kind of like an aged_usr table? |
| 10:17 |
csharp_ |
it looks at actor.usr, action.circulation, action.hold_request, actor.usr_activity, the money.* tables (especially mat xact summary), etc. |
| 10:17 |
csharp_ |
yeah |
| 10:18 |
berick |
could see that being useful |
| 10:18 |
eeevil |
I'm not /super/ crazy about partitioning the live circ/hold tables, but any mat views? 100%, chop them up |
| 10:19 |
csharp_ |
this is what I'm doing: |
| 10:19 |
csharp_ |
https://pastebin.com/ZBBAmtCK |
| 10:19 |
eeevil |
but that's transparent enough that it real should not be baked into stock |
| 10:20 |
eeevil |
(that may sound backwards, but hear me out...) |
| 10:20 |
csharp_ |
eeevil: I agree about not doing anything with stock at this point - just making local decisions (though obvs would pass up anything useful to the community if applicable) |
| 10:20 |
eeevil |
heh +1 |
| 10:22 |
csharp_ |
also maybe window functions in my queries would help (as far as I can tell those are kind of partitioning on the fly?) |
| 10:23 |
eeevil |
ah, so, a key criterium that seems like it should be addressable directly would be an index on expire date (and not worrying about the date() transform in the where clause) |
| 10:24 |
* csharp_ |
looks |
| 10:25 |
csharp_ |
oh - right - I though there was one on there already - I'll try that |
| 10:25 |
csharp_ |
*thought |
| 10:28 |
eeevil |
maybe even use that in a separate CTE that the one called "x" uses. WITH old_users AS (SELECT id FROM actor.usr WHERE expire_date < now() - '4 years'::interval AND create_date $IS_OLD AND profile IN (...)), x as (... FROM old_users LEFT JOIN mmbxs ON (mmbxs.usr=old_users.id AND balance_owed <> '0.00' OR (NOW() - last_payment_ts) < '4 years'::INTERVAL) WHERE mmbxs.id IS NULL JOIN aua ... |
| 10:29 |
csharp_ |
eeevil: nice |
| 10:29 |
csharp_ |
just added the expire_date idx and still getting seq scans, but your suggestion gives me direction - thank you |
| 10:30 |
eeevil |
+1 |
| 10:32 |
csharp_ |
this and a few other "slow query" things have me interested in speeding a lot of things up - this has my wheels turning - it will probably be next week before I have something to really work with - thanks again |
| 10:44 |
Bmagic |
anyone remember the bug about Angular reporter interface not loading your completed reports when you have more than (some magic number) of completed reports? It times out and shows nothing |
| 10:45 |
eeevil |
yes, there was a patch for that |
| 11:01 |
|
mmorgan1 joined #evergreen |
| 12:08 |
csharp_ |
:q |
| 12:08 |
csharp_ |
yeah treatin' IRC like vim like a baws |
| 12:40 |
berick |
@who saw [band] learning vim |
| 12:40 |
pinesol |
eeevil saw Redis Bus learning vim. |
| 12:40 |
berick |
heh |
| 13:11 |
|
mantis joined #evergreen |
| 13:12 |
mantis |
I'm trying to add a Due Date column to the item buckets column picker. This may have been answered before, but if the data doesn't show up, is it because of the path and how that associates with the file? I basically just took it from another file such as t_checkout.tt2 |
| 13:26 |
mantis |
I'm also trying to reference fm_IDL but it seems a lot of these paths don't seem to work well with t_view.tt2 for copy buckets |
| 13:44 |
eeevil |
mantis: you're going to have to do more than just add a grid column. bucket->entry->item->circ link path is not "just the last one, and only if it's open" so the data source will have to be made smarter. |
| 14:37 |
|
mantis joined #evergreen |
| 15:03 |
|
mantis left #evergreen |
| 17:02 |
|
mmorgan left #evergreen |
