Bug #13972
openListing collections by PDH and name can be very slow
Description
Requests to list collections by PDH and name (I think these are coming from a-c-r) sometimes take a very long time (90s+):
{ "method": "GET", "path": "/arvados/v1/collections", "format": "json", "controller": "Arvados::V1::CollectionsController", "action": "index", "status": 200, "duration": 104537.05, "view": 0.33, "db": 104532.55, "request_id": "req-mgxr3wrwt09hytd2jjsc", "client_ipaddr": "10.101.0.7", "client_auth": "eglyx-gj3su-u157eevzet2a0gt", "params": { "alt": "json", "limit": "1", "filters": "[[\"portable_data_hash\", \"=\", \"d41d8cd98f00b204e9800998ecf8427e+0\"], [\"name\", \"like\", \"gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%\"]]" }, "@timestamp": "2018-08-06T17:46:29.775676929Z", "@version": "1", "message": "[200] GET /arvados/v1/collections (Arvados::V1::CollectionsController#index)" }
root@arvados-api-backend-eglyx-01:/var/www/arvados-api/current/log# grep req-mgxr3wrwt09hytd2jjsc production.log [req-mgxr3wrwt09hytd2jjsc] Collection Load (49993.9ms) SELECT collections."uuid", collections."owner_uuid", collections."created_at", collections."modified_by_client_uuid", collections."modified_by_user_uuid", collections."modified_at", collections."name", collections."description", collections."properties", collections."portable_data_hash", collections."replication_desired", collections."replication_confirmed", collections."replication_confirmed_at", collections."storage_classes_desired", collections."storage_classes_confirmed", collections."storage_classes_confirmed_at", collections."delete_at", collections."trash_at", collections."is_trashed" FROM "collections" WHERE (NOT EXISTS(SELECT 1 FROM materialized_permission_view WHERE trashed = 1 AND (collections.owner_uuid = target_uuid)) AND collections.is_trashed = false) AND ((collections.portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0') AND (collections.name like 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%')) ORDER BY collections.modified_at desc, collections.uuid LIMIT 1 OFFSET 0 [req-mgxr3wrwt09hytd2jjsc] (54538.7ms) SELECT COUNT("collections"."id") FROM "collections" WHERE (NOT EXISTS(SELECT 1 FROM materialized_permission_view WHERE trashed = 1 AND (collections.owner_uuid = target_uuid)) AND collections.is_trashed = false) AND ((collections.portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0') AND (collections.name like 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%')) {"method":"GET","path":"/arvados/v1/collections","format":"json","controller":"Arvados::V1::CollectionsController","action":"index","status":200,"duration":104537.05,"view":0.33,"db":104532.55,"request_id":"req-mgxr3wrwt09hytd2jjsc","client_ipaddr":"10.101.0.7","client_auth":"eglyx-gj3su-u157eevzet2a0gt","params":{"alt":"json","limit":"1","filters":"[[\"portable_data_hash\", \"=\", \"d41d8cd98f00b204e9800998ecf8427e+0\"], [\"name\", \"like\", \"gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%\"]]"},"@timestamp":"2018-08-06T17:46:29.775676929Z","@version":"1","message":"[200] GET /arvados/v1/collections (Arvados::V1::CollectionsController#index)"}
Here is how the database explains it:
arvados_api_production=# explain analyze SELECT collections."uuid", collections."owner_uuid", collections."created_at", collections."modified_by_client_uuid", collections."modified_by_user_uuid", collections."modified_at", collections."name", collections."description", collections."properties", collections."portable_data_hash", collections."replication_desired", collections."replication_confirmed", collections."replication_confirmed_at", collections."storage_classes_desired", collections."storage_classes_confirmed", collections."storage_classes_confirmed_at", collections."delete_at", collections."trash_at", collections."is_trashed" FROM "collections" WHERE (NOT EXISTS(SELECT 1 FROM materialized_permission_view WHERE trashed = 1 AND (collections.owner_uuid = target_uuid)) AND collections.is_trashed = false) AND ((collections.portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0') AND (collections.name like 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%')) ORDER BY collections.modified_at desc, collections.uuid LIMIT 1 OFFSET 0; QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Limit (cost=110280.76..110280.76 rows=1 width=894) (actual time=33782.855..33782.855 rows=0 loops=1) -> Sort (cost=110280.76..110280.77 rows=4 width=894) (actual time=33782.855..33782.855 rows=0 loops=1) Sort Key: collections.modified_at DESC, collections.uuid Sort Method: quicksort Memory: 25kB -> Nested Loop Anti Join (cost=1590.26..110280.74 rows=4 width=894) (actual time=33782.847..33782.847 rows=0 loops=1) Join Filter: ((collections.owner_uuid)::text = (materialized_permission_view.target_uuid)::text) -> Bitmap Heap Scan on collections (cost=1590.26..110279.19 rows=5 width=894) (actual time=33782.846..33782.846 rows=0 loops=1) Recheck Cond: ((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) Filter: ((NOT is_trashed) AND ((name)::text ~~ 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%'::text)) Rows Removed by Filter: 65732 Heap Blocks: exact=46329 -> Bitmap Index Scan on index_collections_on_portable_data_hash (cost=0.00..1590.26 rows=61293 width=0) (actual time=352.125..352.125 rows=65732 loops=1) Index Cond: ((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) -> Materialize (cost=0.00..1.48 rows=1 width=28) (never executed) -> Seq Scan on materialized_permission_view (cost=0.00..1.48 rows=1 width=28) (never executed) Filter: (trashed = 1) Planning time: 18.993 ms Execution time: 33782.934 ms (18 rows) arvados_api_production=# explain analyze SELECT COUNT("collections"."id") FROM "collections" WHERE (NOT EXISTS(SELECT 1 FROM materialized_permission_view WHERE trashed = 1 AND (collections.owner_uuid = target_uuid)) AND collections.is_trashed = false) AND ((collections.portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0') AND (collections.name like 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%')); QUERY PLAN --------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Aggregate (cost=110280.74..110280.75 rows=1 width=4) (actual time=36774.915..36774.915 rows=1 loops=1) -> Hash Anti Join (cost=1591.75..110280.73 rows=4 width=4) (actual time=36774.911..36774.911 rows=0 loops=1) Hash Cond: ((collections.owner_uuid)::text = (materialized_permission_view.target_uuid)::text) -> Bitmap Heap Scan on collections (cost=1590.26..110279.19 rows=5 width=32) (actual time=36774.909..36774.909 rows=0 loops=1) Recheck Cond: ((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) Filter: ((NOT is_trashed) AND ((name)::text ~~ 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%'::text)) Rows Removed by Filter: 65732 Heap Blocks: exact=46329 -> Bitmap Index Scan on index_collections_on_portable_data_hash (cost=0.00..1590.26 rows=61293 width=0) (actual time=580.662..580.662 rows=65732 loops=1) Index Cond: ((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) -> Hash (cost=1.48..1.48 rows=1 width=28) (never executed) -> Seq Scan on materialized_permission_view (cost=0.00..1.48 rows=1 width=28) (never executed) Filter: (trashed = 1) Planning time: 0.388 ms Execution time: 36775.002 ms (15 rows)
Updated by Joshua Randall over 6 years ago
arvados_api_production=# select count(*) from collections where portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0'; count ------- 65732 (1 row) arvados_api_production=# select count(*) from collections where portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0' and is_trashed = false; count ------- 17789 (1 row)
Updated by Joshua Randall over 6 years ago
I added another index and this query got 10x faster:
arvados_api_production=# \d index_collections_on_portable_data_hash_is_trashed Index "public.index_collections_on_portable_data_hash_is_trashed" Column | Type | Definition --------------------+------------------------+-------------------- portable_data_hash | character varying(255) | portable_data_hash is_trashed | boolean | is_trashed btree, for table "public.collections" arvados_api_production=# explain analyze SELECT collections."uuid", collections."owner_uuid", collections."created_at", collections."modified_by_client_uuid", collections."modified_by_user_uuid", collections."modified_at", collections."name", collections."description", collections."properties", collections."portable_data_hash", collections."replication_desired", collections."replication_confirmed", collections."replication_confirmed_at", collections."storage_classes_desired", collections."storage_classes_confirmed", collections."storage_classes_confirmed_at", collections."delete_at", collections."trash_at", collections."is_trashed" FROM "collections" WHERE (NOT EXISTS(SELECT 1 FROM materialized_permission_view WHERE trashed = 1 AND (collections.owner_uuid = target_uuid)) AND collections.is_trashed = false) AND ((collections.portable_data_hash = 'd41d8cd98f00b204e9800998ecf8427e+0') AND (collections.name like 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%')) ORDER BY collections.modified_at desc, collections.uuid LIMIT 1 OFFSET 0; QUERY PLAN ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Limit (cost=93374.56..93374.56 rows=1 width=894) (actual time=2944.982..2944.982 rows=0 loops=1) -> Sort (cost=93374.56..93374.57 rows=4 width=894) (actual time=2944.980..2944.980 rows=0 loops=1) Sort Key: collections.modified_at DESC, collections.uuid Sort Method: quicksort Memory: 25kB -> Nested Loop Anti Join (cost=1255.44..93374.54 rows=4 width=894) (actual time=2944.970..2944.970 rows=0 loops=1) Join Filter: ((collections.owner_uuid)::text = (materialized_permission_view.target_uuid)::text) -> Bitmap Heap Scan on collections (cost=1255.44..93372.99 rows=5 width=894) (actual time=2944.969..2944.969 rows=0 loops=1) Recheck Cond: ((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) Filter: ((NOT is_trashed) AND ((name)::text ~~ 'gatk-4.0.0.0-library-cram-to-gvcfs.cwl dependencies%'::text)) Rows Removed by Filter: 17789 Heap Blocks: exact=13978 -> Bitmap Index Scan on index_collections_on_portable_data_hash_is_trashed (cost=0.00..1255.44 rows=51288 width=0) (actual time=21.035..21.035 rows=17789 loops=1) Index Cond: (((portable_data_hash)::text = 'd41d8cd98f00b204e9800998ecf8427e+0'::text) AND (is_trashed = false)) -> Materialize (cost=0.00..1.48 rows=1 width=28) (never executed) -> Seq Scan on materialized_permission_view (cost=0.00..1.48 rows=1 width=28) (never executed) Filter: (trashed = 1) Planning time: 0.704 ms Execution time: 2945.049 ms (18 rows)
Updated by Tom Morris over 6 years ago
- Description updated (diff)
- Target version set to To Be Groomed
I updated the original ticket with the contents of Note-1. I'll delete that note too.
Another thing to note here is that the representation of a 16 byte MD5 hash as a 32 character hexadecimal string has much lower information density than its original representation.
Updated by Tom Morris over 6 years ago
It looks like the Postgres UUID datatype with a B-tree index, may be a more efficient way to go for storing and indexing MD5s:
https://dba.stackexchange.com/questions/115271/what-is-the-optimal-data-type-for-an-md5-field
https://dba.stackexchange.com/questions/102448/how-should-i-index-a-uuid-in-postgres
Also, I'm wondering if the separate COUNT query is really necessary. I guess if both queries are made fast, it's less of an issue, but it makes me suspicious of the anti-pattern of counting things that don't really need counting.
Updated by Peter Amstutz over 3 years ago
- Target version deleted (
To Be Groomed)
Updated by Brett Smith almost 2 years ago
- Related to Bug #14070: API DB needs an index on collections name added