Vps – OpenVZ: shmpages keep increasing

memoryopenvzvps

I have an OpenVZ VPS. However the shmpages always increases to the limit in one or two days. My RAM usage is low (233M without cache). I only run nginx, php-fpm, and mariadb on the VPS. I've asked my VPS provider to increase the limit once (from 13000), but it's still a problem for me. What could be the reason and is there a possible way to fix it? The following is the output of cat /proc/user_beancounters.

Version: 2.5
       uid  resource                     held              maxheld              barrier                limit              failcnt
    39506:  kmemsize                 11328858             17010688  9223372036854775807  9223372036854775807                    0
            lockedpages                     0                    0                  256                  256                    0
            privvmpages                175735               226459  9223372036854775807  9223372036854775807                    0
            shmpages                    19985                20000                20000                20000                 1752
            dummy                           0                    0  9223372036854775807  9223372036854775807                    0
            numproc                        44                  126                  250                  250                    0
            physpages                   78891               101586               361984               361984                    0
            vmguarpages                     0                    0               260144  9223372036854775807                    0
            oomguarpages                37212                44403               260144  9223372036854775807                    0
            numtcpsock                      8                   31                  800                  800                    0
            numflock                        5                   28                  600                  800                    0
            numpty                          1                    1                   47                   47                    0
            numsiginfo                      0                   78                  256                  256                    0
            tcpsndbuf                  139520               721088  9223372036854775807  9223372036854775807                    0
            tcprcvbuf                  131072              4063928  9223372036854775807  9223372036854775807                    0
            othersockbuf                57800               325704  9223372036854775807  9223372036854775807                    0
            dgramrcvbuf                     0                 9248               575488               575488                    0
            numothersock                  109                  135                  900                  900                    0
            dcachesize                3614571              4086400              4086400              4086400                    0
            numfile                       641                 1275                 9600                 9600                    0
            dummy                           0                    0  9223372036854775807  9223372036854775807                    0
            dummy                           0                    0  9223372036854775807  9223372036854775807                    0
            dummy                           0                    0  9223372036854775807  9223372036854775807                    0
            numiptent                      24                   24                  450                  450                    0

EDIT:

df -T output: 

Filesystem        Type     1K-blocks     Used Available Use% Mounted on
devtmpfs          devtmpfs    723968        0    723968   0% /dev
tmpfs             tmpfs       723968        0    723968   0% /dev/shm
tmpfs             tmpfs       723968    13096    710872   2% /run
tmpfs             tmpfs       723968        0    723968   0% /sys/fs/cgroup

Best Answer

A good provider should not be setting the limit for shmpages as low as 20000. Shmpages only affects the resource shortage reaction of applications in your container; it has zero effect on the stability or security of the node itself.

All of my OpenVZ VPS's (from BuyVM) have the shmpages limit set to 9223372036854775807 (basically, unlimited). It's obvious that you are not running out of memory -- your physpages limit is 361984, and the highest amount of memory you've used is less than one-third of that (101586). Same goes with oomguarpages -- your limit is 260144 pages but the maximum memory you have used is 44403 pages, only 17% of what is available to you.

Ask you provider to increase your shmpages limit and barrier to something more sane. If they refuse, I would advise looking elsewhere for your VPS needs. Their limit of 20000 is ridiculous.

Related Topic