Document Type | Troubleshooting
Category | Monitoring/Inspection
Applicable Product Versions | 5SP1FS01, 5SP1FS02, 5SP1FS03, 5SP1FS04, 5SP1FS06, 6FS01, 6FS02, 6FS03, 6FS04, 6FS05, 6FS06, 6FS07, 6FS07PS, 7FS01, 7FS02, 7FS02PS
Document Number | TMOTS019
Issue
When monitoring with ps aux or top, the memory usage may appear larger than the actual memory usage.
Example) The VSZ or RSS of each memory is large, and when all are summed, an inaccurate value that is several times the server memory is displayed.
[tibero1@prolinux1 sql]$ ps aux --sort -rss USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND tibero1 1766 0.0 1.9 3103216 73776 pts/0 Sl 00:01 0:01 tbsvr_FGWP000 -t NORMAL -SVR_SID DB1_node1 (omitted) tibero1 1768 0.0 0.6 2495088 23208 pts/0 Sl 00:01 0:00 tbsvr_PEWP000 -t NORMAL -SVR_SID DB1_node1 tibero1 1773 0.0 0.6 2495088 23052 pts/0 Sl 00:01 0:00 tbsvr_PEWP005 -t NORMAL -SVR_SID DB1_node1 tibero1 1769 0.0 0.6 2495088 23008 pts/0 Sl 00:01 0:00 tbsvr_PEWP001 -t NORMAL -SVR_SID DB1_node1 tibero1 1772 0.0 0.6 2495088 23000 pts/0 Sl 00:01 0:00 tbsvr_PEWP004 -t NORMAL -SVR_SID DB1_node1
Cause
Since Tibero uses shared memory, memory usage displayed by ps or top includes the shared memory area. Therefore, the memory usage shown is the sum of both the actual used memory and the shared memory.
Solutions
In Linux and Solaris environments, you can check the actual memory usage distribution and shared memory using pmap; on AIX, use procmap.
Example of Checking Memory
[tibero1@prolinux1 sql]$ pmap -x 1766 1766: tbsvr_FGWP000 -t NORMAL -SVR_SID DB1_node1 Address Kbytes RSS Dirty Mode Mapping 0000000000400000 152244 26396 0 r-x-- tbsvr 0000000009aad000 220 216 4 r---- tbsvr 0000000009ae4000 3160 2604 96 rw--- tbsvr 0000000009dfa000 884 196 196 rw--- [ anon ] (omitted) 00007fbfeeff7000 32 0 0 ----- zero (omitted) 00007fbff4197000 2097152 32576 32576 rw-s- [ shmid=0x0 ] 00007fc074197000 1776 1424 0 r-x-- libc-2.28.so (omitted) ffffffffff600000 4 0 0 r-x-- [ anon ] ---------------- ------- ------- ------- total kB 3103220 75028 42332
The size 2097152 at address 00007fbff4197000 is shared memory with shmid=0x0. The remainder after subtracting this value from the total is the space actually used separately by the process.
Also, it can be confirmed as below that the shared memory actually exists with id 0 (shmid), the size is the same (bytes=2147483648 - 2097152 KB), and 15 processes are sharing it (nattch 15).
[tibero1@prolinux1 sql]$ ipcs -m ------ Shared Memory Segments -------- key shmid owner perms bytes nattch status 0x1553710c 0 tibero1 640 2147483648 15
CautionPerform this according to the guidance of Tmax Tibero Technical Support.