Re: Out of Memory errors are frustrating as heck! - Mailing list pgsql-performance

OK, I went back through that old thread, and I noticed an early opinion by a certain Peter <pmc at citylink> who said that I should provision some swap space. Since I had plenty of disk and no other option I tried that. And it did some magic. Here this is a steady state now:

top - 14:07:32 up 103 days,  9:57,  5 users,  load average: 1.33, 1.05, 0.54
Tasks: 329 total,   2 running, 117 sleeping,   0 stopped,   0 zombie
%Cpu(s): 31.0 us, 11.4 sy,  0.0 ni, 35.3 id, 22.3 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem :  7910376 total,   120524 free,  2174940 used,  5614912 buff/cache
KiB Swap: 16777212 total, 16777212 free,        0 used.  3239724 avail Mem
 PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND5068 postgres  20   0 4352496   4.0g   2.0g R  76.4 52.6   3:01.39 postgres: postgres integrator [local] INSERT 435 root      20   0       0      0      0 S   4.0  0.0  10:52.38 [kswapd0]

and the nice thing is, the backend server process appears to be bounded at 4GB, so there isn't really a "memory leak". And also, the swap space isn't really being used. This may have to do with these vm. sysctl settings, overcommit, etc.

• vm.overcommit_memory = 2 -- values are
  • 0 -- estimate free memory
  • 1 -- always assume there is enough memory
  • 2 -- no over-commit allocate only inside the following two parameters
• vm.overcommit_kbytes = 0 -- how many kB above swap can be over-committed, EITHER this OR
• vm.overcommit_ratio = 50 -- percent of main memory that can be committed over swap,
  • with 0 swap, that percent can be committed
  • i.e., this of 8 GB, 4 GB are reserved for buffer cache
  • not a good idea probably
  • at least we should allow 75% committed, i.e., 6 GB of 8 GB, leaving
    • 2 GB of buffer cache
    • 2 GB of shared buffers
    • 4 GB of all other memory

I have vm.overcommit_memory = 2, _kbytes = 0, _ratio = 50. So this means with _ratio = 50 I can commit 50% of memory, 4GB and this is exactly what the server process wants. So with little impact on the available buffer cache I am in a fairly good position now. The swap (that in my case I set at 2 x main memory = 16G) serves as a buffer to smooth out this peak usage without ever actually paging.

I suppose even without swap I could have set vm.overcommit_ratio = 75, and I notice now that I already commented this much (the above bullet points are my own notes.)

Anyway, for now, I am good. Thank you very much.

regards,
-Gunther

Hi all, I am connecting to a discussion back from April this year. My data has grown and now I am running into new out of memory situations. Meanwhile the world turned from 11.2 to 11.5 which I just installed only to find the same out of memory error.

Have any of the things discussed and proposed, especially this last one by Tomas Vondra, been applied to the 11 releases? Should I try these older patches from April?

regards,
-Gunther

For what it is worth, this is what I am getting:

TopMemoryContext: 67424 total in 5 blocks; 7184 free (7 chunks); 60240 used pgstat TabStatusArray lookup hash table: 8192 total in 1 blocks; 416 free (0 chunks); 7776 used TopTransactionContext: 8192 total in 1 blocks; 7720 free (1 chunks); 472 used Operator lookup cache: 24576 total in 2 blocks; 10760 free (3 chunks); 13816 used TableSpace cache: 8192 total in 1 blocks; 2096 free (0 chunks); 6096 used Type information cache: 24352 total in 2 blocks; 2624 free (0 chunks); 21728 used RowDescriptionContext: 8192 total in 1 blocks; 6896 free (0 chunks); 1296 used MessageContext: 8388608 total in 11 blocks; 3094872 free (4 chunks); 5293736 used   JoinRelHashTable: 16384 total in 2 blocks; 5576 free (1 chunks); 10808 used Operator class cache: 8192 total in 1 blocks; 560 free (0 chunks); 7632 used smgr relation table: 32768 total in 3 blocks; 12720 free (8 chunks); 20048 used TransactionAbortContext: 32768 total in 1 blocks; 32512 free (0 chunks); 256 used Portal hash: 8192 total in 1 blocks; 560 free (0 chunks); 7632 used TopPortalContext: 8192 total in 1 blocks; 7664 free (0 chunks); 528 used   PortalContext: 1024 total in 1 blocks; 624 free (0 chunks); 400 used:     ExecutorState: 202528536 total in 19 blocks; 433464 free (12 chunks); 202095072 used       HashTableContext: 8192 total in 1 blocks; 7656 free (0 chunks); 536 used         HashBatchContext: 10615104 total in 261 blocks; 7936 free (0 chunks); 10607168 used       HashTableContext: 8192 total in 1 blocks; 7688 free (1 chunks); 504 used         HashBatchContext: 13079304 total in 336 blocks; 7936 free (0 chunks); 13071368 used       TupleSort main: 49208 total in 3 blocks; 8552 free (7 chunks); 40656 used         Caller tuples: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       Subplan HashTable Temp Context: 1024 total in 1 blocks; 768 free (0 chunks); 256 used       Subplan HashTable Context: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       Subplan HashTable Temp Context: 1024 total in 1 blocks; 768 free (0 chunks); 256 used       Subplan HashTable Context: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       Subplan HashTable Temp Context: 1024 total in 1 blocks; 768 free (0 chunks); 256 used       Subplan HashTable Context: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       Subplan HashTable Temp Context: 1024 total in 1 blocks; 768 free (0 chunks); 256 used       Subplan HashTable Context: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       Subplan HashTable Temp Context: 1024 total in 1 blocks; 768 free (0 chunks); 256 used       Subplan HashTable Context: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7360 free (0 chunks); 832 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       ExprContext: 1107296256 total in 142 blocks; 6328 free (101 chunks); 1107289928 used       ExprContext: 8192 total in 1 blocks; 7936 free (0 chunks); 256 used       1 more child contexts containing 8192 total in 1 blocks; 7936 free (0 chunks); 256 used Relcache by OID: 16384 total in 2 blocks; 2472 free (2 chunks); 13912 used CacheMemoryContext: 1113488 total in 14 blocks; 16776 free (0 chunks); 1096712 used   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: docsubjh_sjrcode_ndx   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: docsubjh_sjrclass_ndx   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: docsubjh_scopeiid_ndx   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: docsubjh_dociid_ndx   index info: 4096 total in 3 blocks; 2064 free (2 chunks); 2032 used: role_telecom_idx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: role_addr_fkidx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: role_id_fkidx   index info: 2048 total in 2 blocks; 696 free (1 chunks); 1352 used: role_id_idx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: role_name_fkidx   index info: 4096 total in 3 blocks; 2064 free (2 chunks); 2032 used: entity_telecom_idx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: entity_id_fkidx   index info: 2048 total in 2 blocks; 696 free (1 chunks); 1352 used: entity_id_idx   index info: 2048 total in 2 blocks; 624 free (1 chunks); 1424 used: entity_det_code_idx   index info: 4096 total in 3 blocks; 2016 free (2 chunks); 2080 used: entity_code_nodash_idx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: entity_pkey   index info: 2048 total in 2 blocks; 680 free (1 chunks); 1368 used: connect_rule_pkey   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: role_context_idx   index info: 2048 total in 2 blocks; 640 free (2 chunks); 1408 used: role_partitions   index info: 2048 total in 2 blocks; 640 free (2 chunks); 1408 used: role_scoper_idx   index info: 2048 total in 2 blocks; 640 free (2 chunks); 1408 used: role_player_idx   index info: 2048 total in 2 blocks; 968 free (1 chunks); 1080 used: role__pkey   index info: 2048 total in 2 blocks; 680 free (1 chunks); 1368 used: pg_toast_2619_index   index info: 2048 total in 2 blocks; 592 free (1 chunks); 1456 used: pg_constraint_conrelid_contypid_conname_index   index info: 2048 total in 2 blocks; 624 free (1 chunks); 1424 used: participation_act_idx   index info: 2048 total in 2 blocks; 624 free (1 chunks); 1424 used: participation_role_idx   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: participation_pkey   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_statistic_ext_relid_index   index info: 2048 total in 2 blocks; 624 free (1 chunks); 1424 used: doc_ndx_internaiddoctype   index info: 2048 total in 2 blocks; 680 free (1 chunks); 1368 used: pg_toast_2618_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_index_indrelid_index   relation rules: 827392 total in 104 blocks; 2400 free (1 chunks); 824992 used: v_documentsubjecthistory   index info: 2048 total in 2 blocks; 648 free (2 chunks); 1400 used: pg_db_role_setting_databaseid_rol_index   index info: 2048 total in 2 blocks; 624 free (2 chunks); 1424 used: pg_opclass_am_name_nsp_index   index info: 1024 total in 1 blocks; 16 free (0 chunks); 1008 used: pg_foreign_data_wrapper_name_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_enum_oid_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_class_relname_nsp_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_foreign_server_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_publication_pubname_index   index info: 2048 total in 2 blocks; 592 free (3 chunks); 1456 used: pg_statistic_relid_att_inh_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_cast_source_target_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_language_name_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_transform_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_collation_oid_index   index info: 3072 total in 2 blocks; 1136 free (2 chunks); 1936 used: pg_amop_fam_strat_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_index_indexrelid_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_ts_template_tmplname_index   index info: 2048 total in 2 blocks; 704 free (3 chunks); 1344 used: pg_ts_config_map_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_opclass_oid_index   index info: 1024 total in 1 blocks; 16 free (0 chunks); 1008 used: pg_foreign_data_wrapper_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_event_trigger_evtname_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_statistic_ext_name_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_publication_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_ts_dict_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_event_trigger_oid_index   index info: 3072 total in 2 blocks; 1216 free (3 chunks); 1856 used: pg_conversion_default_index   index info: 3072 total in 2 blocks; 1136 free (2 chunks); 1936 used: pg_operator_oprname_l_r_n_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_trigger_tgrelid_tgname_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_enum_typid_label_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_ts_config_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_user_mapping_oid_index   index info: 2048 total in 2 blocks; 704 free (3 chunks); 1344 used: pg_opfamily_am_name_nsp_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_foreign_table_relid_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_type_oid_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_aggregate_fnoid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_constraint_oid_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_rewrite_rel_rulename_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_ts_parser_prsname_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_ts_config_cfgname_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_ts_parser_oid_index   index info: 2048 total in 2 blocks; 728 free (1 chunks); 1320 used: pg_publication_rel_prrelid_prpubid_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_operator_oid_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_namespace_nspname_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_ts_template_oid_index   index info: 2048 total in 2 blocks; 624 free (2 chunks); 1424 used: pg_amop_opr_fam_index   index info: 2048 total in 2 blocks; 672 free (3 chunks); 1376 used: pg_default_acl_role_nsp_obj_index   index info: 2048 total in 2 blocks; 704 free (3 chunks); 1344 used: pg_collation_name_enc_nsp_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_publication_rel_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_range_rngtypid_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_ts_dict_dictname_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_type_typname_nsp_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_opfamily_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_statistic_ext_oid_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_class_oid_index   index info: 2048 total in 2 blocks; 624 free (2 chunks); 1424 used: pg_proc_proname_args_nsp_index   index info: 1024 total in 1 blocks; 16 free (0 chunks); 1008 used: pg_partitioned_table_partrelid_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_transform_type_lang_index   index info: 2048 total in 2 blocks; 680 free (2 chunks); 1368 used: pg_attribute_relid_attnum_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_proc_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_language_oid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_namespace_oid_index   index info: 3072 total in 2 blocks; 1136 free (2 chunks); 1936 used: pg_amproc_fam_proc_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_foreign_server_name_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_attribute_relid_attnam_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_conversion_oid_index   index info: 2048 total in 2 blocks; 728 free (1 chunks); 1320 used: pg_user_mapping_user_server_index   index info: 2048 total in 2 blocks; 728 free (1 chunks); 1320 used: pg_subscription_rel_srrelid_srsubid_index   index info: 1024 total in 1 blocks; 48 free (0 chunks); 976 used: pg_sequence_seqrelid_index   index info: 2048 total in 2 blocks; 760 free (2 chunks); 1288 used: pg_conversion_name_nsp_index   index info: 2048 total in 2 blocks; 952 free (1 chunks); 1096 used: pg_authid_oid_index   index info: 2048 total in 2 blocks; 728 free (1 chunks); 1320 used: pg_auth_members_member_role_index   10 more child contexts containing 17408 total in 17 blocks; 6080 free (10 chunks); 11328 used WAL record construction: 49768 total in 2 blocks; 6368 free (0 chunks); 43400 used PrivateRefCount: 8192 total in 1 blocks; 2624 free (0 chunks); 5568 used MdSmgr: 8192 total in 1 blocks; 6408 free (0 chunks); 1784 used LOCALLOCK hash: 16384 total in 2 blocks; 4600 free (2 chunks); 11784 used Timezones: 104120 total in 2 blocks; 2624 free (0 chunks); 101496 used ErrorContext: 8192 total in 1 blocks; 7936 free (4 chunks); 256 used
Grand total: 1345345736 bytes in 1209 blocks; 4529600 free (270 chunks); 1340816136 used

On 4/28/2019 10:19, Tomas Vondra wrote:

On Wed, Apr 24, 2019 at 02:36:33AM +0200, Tomas Vondra wrote:

...

I still think the idea with an "overflow batch" is worth considering,
because it'd allow us to keep the memory usage within work_mem. And
after getting familiar with the hash join code again (haven't messed
with it since 9.5 or so) I think it should not be all that difficult.
I'll give it a try over the weekend if I get bored for a while.

OK, so I took a stab at this, and overall it seems to be workable. The
patches I have are nowhere near committable, but I think the approach
works fairly well - the memory is kept in check, and the performance is
comparable to the "ballancing" approach tested before.

To explain it a bit, the idea is that we can compute how many BufFile
structures we can keep in memory - we can't use more than work_mem/2 for
that, because then we'd mostly eliminate space for the actual data. For
example with 4MB, we know we can keep 128 batches - we need 128 for
outer and inner side, so 256 in total, and 256*8kB = 2MB.

And then, we just increase the number of batches but instead of adding
the BufFile entries, we split batches into slices that we can keep in
memory (say, the 128 batches). And we keep BufFiles for the current one
and an "overflow file" for the other slices. After processing a slice,
we simply switch to the next one, and use the overflow file as a temp
file for the first batch - we redistribute it into the other batches in
the slice and another overflow file.

That's what the v3 patch (named 'single overflow file') does. I does
work, but unfortunately it significantly inflates the amount of data
written to temporary files. Assume we need e.g. 1024 batches, but only
128 fit into memory. That means we'll need 8 slices, and during the
first pass we'll handle 1/8 of the data and write 7/8 to the overflow
file.  Then after processing the slice and switching to the next one, we
repeat this dance - 1/8 gets processed, 6/8 written to another overflow
file. So essentially we "forward" about

   7/8 + 6/8 + 5/8 + ... + 1/8 = 28/8 = 3.5

of data between slices, and we need to re-shuffle data in each slice,
which amounts to additional 1x data. That's pretty significant overhead,
as will be clear from the measurements I'll present shortly.

But luckily, there's a simple solution to this - instead of writing the
data into a single overflow file, we can create one overflow file for
each slice. That will leave us with the ~1x of additional writes when
distributing data into batches in the current slice, but it eliminates
the main source of write amplification - awalanche-like forwarding of
data between slices.

This relaxes the memory limit a bit again, because we can't really keep
the number of overflow files constrained by work_mem, but we should only
need few of them (much less than when adding one file per batch right
away). For example with 128 in-memory batches, this reduces the amount
of necessary memory 128x.

And this is what v4 (per-slice overflow file) does, pretty much.


Two more comments, regarding memory accounting in previous patches. It
was a bit broken, because we actually need 2x the number of BufFiles. We
needed nbatch files for outer side and nbatch files for inner side, but
we only considered one of those - both when deciding when to increase
the number of batches / increase spaceAllowed, and when reporting the
memory usage. So with large number of batches the reported amount of
used memory was roughly 1/2 of the actual value :-/

The memory accounting was a bit bogus for another reason - spaceUsed
simply tracks the amount of memory for hash table contents. But at the
end we were simply adding the current space for BufFile stuff, ignoring
the fact that that's likely much larger than when the spacePeak value
got stored. For example we might have kept early spaceUsed when it was
almost work_mem, and then added the final large BufFile allocation.

I've fixed both issues in the patches attached to this message. It does
not make a huge difference in practice, but it makes it easier to
compare values between patches.


Now, some test results - I've repeated the simple test with uniform data
set, which is pretty much ideal for hash joins (no unexlectedly large
batches that can't be split, etc.). I've done this with 1M, 5M, 10M, 25M
and 50M rows in the large table (which gets picked for the "hash" side),
and measured how much memory gets used, how many batches, how long it
takes and how much data gets written to temp files.

See the hashjoin-test.sh script for more details.

So, here are the results with work_mem = 4MB (so the number of in-memory
batches for the last two entries is 128). The columns are:

* nbatch - the final number of batches
* memory - memory usage, as reported by explain analyze
* time - duration of the query (without explain analyze) in seconds
* size - size of the large table
* temp - amount of data written to temp files
* amplif - write amplification (temp / size)


 1M rows
 ===================================================================
                 nbatch  memory   time  size (MB)  temp (MB)  amplif
 -------------------------------------------------------------------
 master             256    7681    3.3        730        899    1.23
 rebalance          256    7711    3.3        730        884    1.21
 single file       1024    4161    7.2        730       3168    4.34
 per-slice file    1024    4161    4.7        730       1653    2.26


 5M rows
 ===================================================================
                 nbatch  memory   time  size (MB)  temp (MB)  amplif
 -------------------------------------------------------------------
 master            2048   36353     22       3652       5276    1.44
 rebalance          512   16515     18       3652       4169    1.14
 single file       4096    4353    156       3652      53897   14.76
 per-slice file    4096    4353     28       3652       8106    2.21


 10M rows
 ===================================================================
                 nbatch  memory   time  size (MB)  temp (MB)  amplif
 -------------------------------------------------------------------
 master            4096   69121     61       7303      10556    1.45
 rebalance          512   24326     46       7303       7405    1.01
 single file       8192    4636    762       7303     211234   28.92
 per-slice file    8192    4636     65       7303      16278    2.23


 25M rows
 ===================================================================
                 nbatch  memory   time  size (MB)  temp (MB)  amplif
 -------------------------------------------------------------------
 master            8192  134657    190       7303      24279    1.33
 rebalance         1024   36611    158       7303      20024    1.10
 single file      16384    6011   4054       7303    1046174   57.32
 per-slice file   16384    6011    207       7303      39073    2.14


 50M rows
 ===================================================================
                 nbatch  memory   time  size (MB)  temp (MB)  amplif
 -------------------------------------------------------------------
 master           16384  265729    531      36500      48519   1.33
 rebalance         2048   53241    447      36500      48077   1.32
 single file          -       -      -      36500          -      -
 per-slice file   32768    8125    451      36500      78662   2.16


From those numbers it's pretty clear that per-slice overflow file does
by far the best job in enforcing work_mem and minimizing the amount of
data spilled to temp files. It does write a bit more data than both
master and the simple rebalancing, but that's the cost for enforcing
work_mem more strictly. It's generally a bit slower than those two
approaches, although on the largest scale it's actually a bit faster
than master. I think that's pretty acceptable, considering this is meant
to address extreme underestimates where we currently just eat memory.

The case with single overflow file performs rather poorly - I haven't
even collected data from the largest scale, but considering it spilled
1TB of temp files with a dataset half the size, that's not an issue.
(Note that this does not mean it needs 1TB of temp space, those writes
are spread over time and the files are created/closed as we go. The
system only has ~100GB of free disk space.)


Gunther, could you try the v2 and v4 patches on your data set? That
would be an interesting data point, I think.


regards