Thread: Specific query taking time to process

There is a specific search query I am running to get list of Documents and their metadata from several table in the DB.

We are running Postgres 9.6.9 on Amazon RDS (db.m5.4xlarge instance)

Our current DB consists of 500GB of data and indexes. Most of the rows in table are consist of 454,078,915

With the fresh DB with the restore of the DATA without any indexes Search query performs relatively quick and most of the time its less than a second. 

But after 3 weeks of use of the DB it sudenly started to slowdown only for this perticular query and it takes 20+ seconds to respond. If I do a restore the DB again then it continues to work fine and the symptom pops out after 3 weeks time. 

You haven't been quite clear on the situation and your use case, but assuming this table has 454 million rows and experiences updates/deletes then this sounds like you may be having problems with autovacuum. Have you customized parameters to ensure it is running more frequently than default? How are you doing those data restores? Perhaps that process is cleaning up the accumulated bloat and you can run fine again for a while. Check pg_stat_user_tables for the last (auto)vacuum that ran, assuming you didn't just restore again and are expecting the issue to occur again soon.

On Mon, Dec 9, 2019 at 3:39 PM Fahiz Mohamed <fahiz@netwidz.com> wrote:

I ran "explain analyse" on this query and I got following result. (We have 2 identical DB instances and they consist of same data. Instane 1 took 20+ second to process and instance 2 took less than a second)

Instance 1: (This is used by regular User - More than 600,000 request a day) - The result is same even when there is no user in the server.

     -> Bitmap Heap Scan on alf_node node (cost=995009.97..3303978.85 rows=4565737 width=8) (actual time=3304.419..20465.551 rows=41109751 loops=1)
        Recheck Cond: ((store_id = 6) AND (type_qname_id = 240))
        Rows Removed by Index Recheck: 54239131
        Filter: (NOT (hashed SubPlan 1))
        Rows Removed by Filter: 2816
        Heap Blocks: exact=24301 lossy=1875383

 Planning time: 0.639 ms
 Execution time: 22946.036 ms

https://paquier.xyz/postgresql-2/postgres-9-4-feature-highlight-lossyexact-pages-for-bitmap-heap-scan/ 

That seems like a lot of lossy blocks. As I understand it, that means the system didn't have enough work_mem to fit all the references to the individual rows which perhaps isn't surprising when it estimates it needs 4.5 million rows and ends up with 41 million.

Do both DB instances have the same data? I ask because the two plans are rather different which makes me think that statistics about the data are not very similar. Are both configured the same, particularly for shared_buffers and work_mem, as well as the various planning cost parameters like random_page cost? If you can provide these plans again with explain( analyze, buffers ) this time? Did you check on the last time autovacuum ran in pg_stat_user_tables?

On Wed, Dec 11, 2019 at 5:21 PM Fahiz Mohamed <fahiz@netwidz.com> wrote:

There is a slight different in both instance’s data. Inastanbce 1 contains latest data and instance 2 consists of data which is 3 weeks older than instance 1. 

In knowing where to look for differences in performance, there is a big difference between them being identical, and being generally similar, but not identical.

I hope the above data difference can make a drastic difference. Please correct me if I am wrong.

They are similar in scale, but we know there is a big difference in distribution of some values.  For example, we still know the slow plan has 4697 rows in aspect_1 where qname_id = 251, while the other plan has 85 rows in aspect_1 meeting that same criterion. That is a big difference, and it is real difference in the data, not just a difference in planning or estimation.  Is this difference driving the difference in plan choice?  Probably not (plan choice is driven by estimated rows, not actual, and estimates are quite similar), but it does demonstrate the data is quite different between the two systems when you look under the hood.  It is likely that there are other, similar differences in the distribution of particular values which is driving the difference in plans.  It is just that we can't see those differences, because the EXPLAIN ANALYZE only reports on the plan it ran, not other plans it could have ran but didn't.

Your query is now using the index named  idx_alf_node_tqn in a way which is equally unproductive as the previous use of  idx_alf_node_mdq was.  It looks like they have the same columns, just in a different order.  My previous advice to try  "type_qname_id+0 = 240" should still apply.

If you can't get that to work, then another avenue is to run "explain (analyze, buffers) select count(*) from alf_node where (type_qname_id = 240) AND (store_id = 6)" on both instances.

I did execute vacuum manually and I noticed the below in the output

"INFO: vacuuming "public.alf_node_aspects"

INFO: "alf_node_aspects": found 0 removable, 150264654 nonremovable row versions in 812242 pages

DETAIL: 0 dead row versions cannot be removed yet.

CPU 13.53s/33.35u sec elapsed 77.88 sec.

I'm not really sure what that means.  I certainly would not have expected 0 removable. There should have been some prior output, something like:

INFO:  scanned index "fk_alf_nasp_qn" to remove 500000 row versions  

 

It could be that autovacuum had already gotten around to vacuuming after your initial email but before you did the above, meaning there was not much for your manual to do.

But you can see that the vacuum did have an effect, by comparing these lines (despite them finding about same number of rows)

Heap Blocks: exact=40765

Heap Blocks: exact=1774 

It wasn't all that large of an effect in this case, but it is still something worth fixing.

Cheers,

Jeff

I have read through your comments so far – and they tally with the checks we have been making to an extent – thanks for them.

We now only need 1 environment as we  can replicate the performance problem on a copy of live – snapshot/restore from AWS of live. We now have a vacuum analyse running every night on the 3 tables in question on live – to eliminate bloat and inaccurate stats as the root of the problem.

We can flip performance based on setting work_mem. For example, currently work_mem up to an including 5069Kb the performance of the query is well under 1 second – Upping work_mem just 1kb to 5097Kb then changes the query plan and performance is a lot worse.

Fresh snapshot and restore.

5096kb – plan –

Nested Loop Semi Join  (cost=3494.04..184505785.24 rows=1 width=8) (actual time=6.404..130.145 rows=212 loops=1)

   Buffers: shared hit=7369

   ->  Nested Loop  (cost=3493.47..184505754.56 rows=36 width=16) (actual time=6.394..129.351 rows=212 loops=1)

         Buffers: shared hit=6318

         ->  HashAggregate  (cost=72.32..81.03 rows=871 width=8) (actual time=0.123..0.186 rows=228 loops=1)

               Group Key: prop.node_id

               Buffers: shared hit=99

               ->  Index Only Scan using idx_alf_nprop_s on alf_node_properties prop  (cost=0.70..70.14 rows=872 width=8) (actual time=0.025..0.086 rows=228 loops=1)

                     Index Cond: ((qname_id = '242'::bigint) AND (string_value = 'S434071'::text))

                     Heap Fetches: 2

                     Buffers: shared hit=99

         ->  Index Only Scan using idx_alf_node_tqn on alf_node node  (cost=3421.15..211831.99 rows=1 width=8) (actual time=0.566..0.566 rows=1 loops=228)

               Index Cond: ((type_qname_id = 240) AND (store_id = 6) AND (id = prop.node_id))

               Filter: (NOT (SubPlan 1))

               Heap Fetches: 13

               Buffers: shared hit=6219

               SubPlan 1

                 ->  Materialize  (cost=3420.59..419826.13 rows=163099 width=8) (actual time=0.007..0.246 rows=4909 loops=212)

                       Buffers: shared hit=5092

                       ->  Bitmap Heap Scan on alf_node_aspects aspect_1  (cost=3420.59..418372.63 rows=163099 width=8) (actual time=1.402..5.243 rows=4909 loops=1)

                             Recheck Cond: (qname_id = 251)

                             Heap Blocks: exact=4801

                             Buffers: shared hit=5092

                             ->  Bitmap Index Scan on fk_alf_nasp_qn  (cost=0.00..3379.81 rows=163099 width=0) (actual time=0.850..0.850 rows=7741 loops=1)

                                   Index Cond: (qname_id = 251)

                                   Buffers: shared hit=291

   ->  Index Only Scan using alf_node_aspects_pkey on alf_node_aspects aspect  (cost=0.57..4.70 rows=15 width=8) (actual time=0.003..0.003 rows=1 loops=212)

         Index Cond: ((node_id = node.id) AND (qname_id = 245))

         Heap Fetches: 15

         Buffers: shared hit=1051

 Planning time: 0.624 ms

 Execution time: 130.222 ms

(32 rows)

5097Kb – plan –

Nested Loop Semi Join  (cost=1019128.07..3400161.81 rows=1 width=8) (actual time=4832.639..32783.503 rows=212 loops=1)

   Buffers: shared hit=565 read=2191862

   ->  Hash Semi Join  (cost=1019127.50..3400131.13 rows=36 width=16) (actual time=4832.599..32779.613 rows=212 loops=1)

         Hash Cond: (node.id = prop.node_id)

         Buffers: shared hit=25 read=2191476

         ->  Bitmap Heap Scan on alf_node node  (cost=1019046.46..3388792.78 rows=4273414 width=8) (actual time=4219.440..29678.682 rows=41976707 loops=1)

               Recheck Cond: ((store_id = 6) AND (type_qname_id = 240))

               Rows Removed by Index Recheck: 58872880

               Filter: (NOT (hashed SubPlan 1))

               Rows Removed by Filter: 2453

               Heap Blocks: exact=32899 lossy=1939310

               Buffers: shared read=2191402

               ->  Bitmap Index Scan on idx_alf_node_mdq  (cost=0.00..599197.73 rows=19566916 width=0) (actual time=4186.449..4186.449 rows=41990879 loops=1)

                     Index Cond: ((store_id = 6) AND (type_qname_id = 240))

                     Buffers: shared read=214101

               SubPlan 1

                 ->  Bitmap Heap Scan on alf_node_aspects aspect_1  (cost=3420.59..418372.63 rows=163099 width=8) (actual time=2.635..21.981 rows=4909 loops=1)

                       Recheck Cond: (qname_id = 251)

                       Heap Blocks: exact=4801

                       Buffers: shared read=5092

                       ->  Bitmap Index Scan on fk_alf_nasp_qn  (cost=0.00..3379.81 rows=163099 width=0) (actual time=2.016..2.016 rows=7741 loops=1)

                             Index Cond: (qname_id = 251)

                             Buffers: shared read=291

         ->  Hash  (cost=70.14..70.14 rows=872 width=8) (actual time=0.357..0.357 rows=228 loops=1)

               Buckets: 1024  Batches: 1  Memory Usage: 17kB

               Buffers: shared hit=25 read=74

               ->  Index Only Scan using idx_alf_nprop_s on alf_node_properties prop  (cost=0.70..70.14 rows=872 width=8) (actual time=0.047..0.325 rows=228 loops=1)

                     Index Cond: ((qname_id = '242'::bigint) AND (string_value = 'S434071'::text))

                     Heap Fetches: 2

                     Buffers: shared hit=25 read=74

   ->  Index Only Scan using alf_node_aspects_pkey on alf_node_aspects aspect  (cost=0.57..4.70 rows=15 width=8) (actual time=0.016..0.016 rows=1 loops=212)

         Index Cond: ((node_id = node.id) AND (qname_id = 245))

         Heap Fetches: 15

         Buffers: shared hit=540 read=386

 Planning time: 0.821 ms

 Execution time: 32783.609 ms

(36 rows)

Note the second plan is not even using any index on alf_node initially.