ERROR: could not resize shared memory segment...No space left ondevice

Поиск
Список
Период
Сортировка
От Justin Pryzby
Тема ERROR: could not resize shared memory segment...No space left ondevice
Дата
Msg-id 20191216184906.GA2082@telsasoft.com
обсуждение исходный текст
Ответы Re: ERROR: could not resize shared memory segment...No space lefton device  (Tomas Vondra <tomas.vondra@2ndquadrant.com>)
Список pgsql-hackers
A customer's report query hit this error.
ERROR:  could not resize shared memory segment "/PostgreSQL.2011322019" to 134217728 bytes: No space left on device

I found:

https://www.postgresql.org/message-id/flat/CAEepm%3D2D_JGb8X%3DLa-0PX9C8dBX9%3Dj9wY%2By1-zDWkcJu0%3DBQbA%40mail.gmail.com

work_mem | 128MB
dynamic_shared_memory_type | posix
version | PostgreSQL 12.1 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23), 64-bit
Running centos 6.9 / linux 2.6.32-696.23.1.el6.x86_64

$ free -m
             total       used       free     shared    buffers     cached
Mem:          7871       7223        648       1531          5       1988
-/+ buffers/cache:       5229       2642
Swap:         4095       2088       2007

$ mount | grep /dev/shm
tmpfs on /dev/shm type tmpfs (rw)

$ du -hs /dev/shm
0       /dev/shm

$ df /dev/shm
Filesystem     1K-blocks  Used Available Use% Mounted on
tmpfs            4030272    24   4030248   1% /dev/shm

Later, I see:
$ df -h /dev/shm
Filesystem      Size  Used Avail Use% Mounted on
tmpfs           3.9G  3.3G  601M  85% /dev/shm

I can reproduce the error running a single instance of the query.

The query plan is 1300 lines long, and involves 482 "Scan" nodes on a table
which currently has 93 partitions, and for which current partitions are
"daily".  I believe I repartitioned its history earlier this year to "monthly",
probably to avoid "OOM with many sorts", as reported here:
https://www.postgresql.org/message-id/20190708164401.GA22387%40telsasoft.com

$ grep Scan tmp/sql-`date +%F`.1.ex |sed 's/^ *//; s/ on .*//' |sort |uniq -c |sort -nr 
    227 ->  Parallel Bitmap Heap Scan
    227 ->  Bitmap Index Scan
     14 ->  Parallel Seq Scan
      9 ->  Seq Scan
      2 ->  Subquery Scan
      2 ->  Index Scan using sites_pkey
      1 Subquery Scan

There are total of 10 "Workers Planned":
grep -o 'Worker.*' tmp/sql-`date +%F`.1.ex
Workers Planned: 2
Workers Planned: 2
Workers Planned: 2
Workers Planned: 2
Workers Planned: 2

I will plan to repartition again to month granularity unless someone wants to
collect further information or suggest a better solution.

(gdb) bt
#0  pg_re_throw () at elog.c:1717
#1  0x0000000000886194 in errfinish (dummy=<value optimized out>) at elog.c:464
#2  0x0000000000749453 in dsm_impl_posix (op=<value optimized out>, 
    handle=<value optimized out>, request_size=<value optimized out>, 
    impl_private=<value optimized out>, mapped_address=<value optimized out>, 
    mapped_size=<value optimized out>, elevel=20) at dsm_impl.c:283
#3  dsm_impl_op (op=<value optimized out>, handle=<value optimized out>, 
    request_size=<value optimized out>, impl_private=<value optimized out>, 
    mapped_address=<value optimized out>, mapped_size=<value optimized out>, 
    elevel=20) at dsm_impl.c:170
#4  0x000000000074a7c8 in dsm_create (size=100868096, flags=0) at dsm.c:459
#5  0x00000000008a94a6 in make_new_segment (area=0x1d70208, 
    requested_pages=<value optimized out>) at dsa.c:2156
#6  0x00000000008aa47a in dsa_allocate_extended (area=0x1d70208, 
    size=100663304, flags=5) at dsa.c:712
#7  0x0000000000670b3f in pagetable_allocate (pagetable=<value optimized out>, 
    size=<value optimized out>) at tidbitmap.c:1511
#8  0x000000000067200c in pagetable_grow (tbm=0x7f82274da8e8, pageno=906296)
    at ../../../src/include/lib/simplehash.h:405
#9  pagetable_insert (tbm=0x7f82274da8e8, pageno=906296)
    at ../../../src/include/lib/simplehash.h:530
#10 tbm_get_pageentry (tbm=0x7f82274da8e8, pageno=906296) at tidbitmap.c:1225
#11 0x00000000006724a0 in tbm_add_tuples (tbm=0x7f82274da8e8, 
    tids=<value optimized out>, ntids=1, recheck=false) at tidbitmap.c:405
#12 0x00000000004d7f1f in btgetbitmap (scan=0x1d7d948, tbm=0x7f82274da8e8)
    at nbtree.c:334
#13 0x00000000004d103a in index_getbitmap (scan=0x1d7d948, 
    bitmap=<value optimized out>) at indexam.c:665
#14 0x00000000006323d8 in MultiExecBitmapIndexScan (node=0x1dcbdb8)
    at nodeBitmapIndexscan.c:105
#15 0x00000000006317f4 in BitmapHeapNext (node=0x1d8a030)
    at nodeBitmapHeapscan.c:141
#16 0x000000000062405c in ExecScanFetch (node=0x1d8a030, 
    accessMtd=0x6316d0 <BitmapHeapNext>, 
    recheckMtd=0x631440 <BitmapHeapRecheck>) at execScan.c:133
#17 ExecScan (node=0x1d8a030, accessMtd=0x6316d0 <BitmapHeapNext>, 
    recheckMtd=0x631440 <BitmapHeapRecheck>) at execScan.c:200
#18 0x0000000000622900 in ExecProcNodeInstr (node=0x1d8a030)
    at execProcnode.c:461
#19 0x000000000062c66f in ExecProcNode (pstate=0x1d7ad70)
    at ../../../src/include/executor/executor.h:239
#20 ExecAppend (pstate=0x1d7ad70) at nodeAppend.c:292
#21 0x0000000000622900 in ExecProcNodeInstr (node=0x1d7ad70)
    at execProcnode.c:461
#22 0x0000000000637da2 in ExecProcNode (pstate=0x1d7a630)
    at ../../../src/include/executor/executor.h:239
#23 ExecHashJoinOuterGetTuple (pstate=0x1d7a630) at nodeHashjoin.c:833
#24 ExecHashJoinImpl (pstate=0x1d7a630) at nodeHashjoin.c:356
#25 ExecHashJoin (pstate=0x1d7a630) at nodeHashjoin.c:572
#26 0x0000000000622900 in ExecProcNodeInstr (node=0x1d7a630)
    at execProcnode.c:461
#27 0x0000000000637da2 in ExecProcNode (pstate=0x1d7bff0)
    at ../../../src/include/executor/executor.h:239
#28 ExecHashJoinOuterGetTuple (pstate=0x1d7bff0) at nodeHashjoin.c:833
#29 ExecHashJoinImpl (pstate=0x1d7bff0) at nodeHashjoin.c:356
#30 ExecHashJoin (pstate=0x1d7bff0) at nodeHashjoin.c:572
#31 0x0000000000622900 in ExecProcNodeInstr (node=0x1d7bff0)
    at execProcnode.c:461
#32 0x000000000061eac7 in ExecProcNode (queryDesc=0x7f8228b72198, 
    direction=<value optimized out>, count=0, execute_once=240)
    at ../../../src/include/executor/executor.h:239
#33 ExecutePlan (queryDesc=0x7f8228b72198, direction=<value optimized out>, 
    count=0, execute_once=240) at execMain.c:1646
#34 standard_ExecutorRun (queryDesc=0x7f8228b72198, 
    direction=<value optimized out>, count=0, execute_once=240)
    at execMain.c:364
#35 0x00007f8229aa7878 in pgss_ExecutorRun (queryDesc=0x7f8228b72198, 
    direction=ForwardScanDirection, count=0, execute_once=true)
    at pg_stat_statements.c:893
#36 0x00007f8228f8d9ad in explain_ExecutorRun (queryDesc=0x7f8228b72198, 
    direction=ForwardScanDirection, count=0, execute_once=true)
    at auto_explain.c:320
#37 0x000000000061f0ce in ParallelQueryMain (seg=0x1c8d3b8, toc=0x7f82291f1000)
    at execParallel.c:1399
#38 0x00000000004f7daf in ParallelWorkerMain (main_arg=<value optimized out>)
    at parallel.c:1431
#39 0x00000000006eb2e0 in StartBackgroundWorker () at bgworker.c:834
#40 0x00000000006f52ac in do_start_bgworker () at postmaster.c:5770
#41 maybe_start_bgworkers () at postmaster.c:5996
#42 0x00000000006f867d in sigusr1_handler (
    postgres_signal_arg=<value optimized out>) at postmaster.c:5167
#43 <signal handler called>
#44 0x0000003049ae1603 in __select_nocancel () from /lib64/libc.so.6
#45 0x00000000006f9d43 in ServerLoop (argc=<value optimized out>, 
    argv=<value optimized out>) at postmaster.c:1668
#46 PostmasterMain (argc=<value optimized out>, argv=<value optimized out>)
    at postmaster.c:1377
#47 0x000000000066a6b0 in main (argc=3, argv=0x1c5b950) at main.c:228

bt f:
#2  0x0000000000749453 in dsm_impl_posix (op=<value optimized out>, handle=<value optimized out>, request_size=<value
optimizedout>, impl_private=<value optimized out>, mapped_address=<value optimized out>, mapped_size=<value optimized
out>,elevel=20) at dsm_impl.c:283
 
        save_errno = <value optimized out>
        st = {st_dev = 26, st_ino = 0, st_nlink = 33554432, st_mode = 4096, st_uid = 0, st_gid = 65536, __pad0 = 0,
st_rdev= 8975118, st_size = 496105863, st_blksize = 8974012, st_blocks = 140199374652304, st_atim = {tv_sec = 7593264,
tv_nsec= 20}, st_mtim = {tv_sec = 85899345920, tv_nsec = 140198902917632}, 
 
          st_ctim = {tv_sec = 44417024, tv_nsec = 289}, __unused = {7593264, 0, 140724603453440}}
        flags = <value optimized out>
        fd = <value optimized out>
        name =
"/PostgreSQL.1648263397\000\000\024\242m\371\000\000\000\000\000\061w3\202\177\000\000\000\000\000\001\000\000\000\000pe\363\001\000\000\000\000\360\271\305\001\000\000\000"
        address = <value optimized out>

Thanks,
Justin



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Ranier Vilela
Дата:
Сообщение: Windows port minor fixes
Следующее
От: Peter Eisentraut
Дата:
Сообщение: polymorphic table functions light