Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm investigating the virtual memory configuration of a Debian based Amazon EC2 instance, and as my background isn't in system admin, I'm slightly confused by what I'm seeing.

We're using MongoDB, and the monitoring server we have indicates that the Mongo process is using about 20GB of swap space virtual memory, however I can't figure out where this is located on the server.

As far as I can tell from using the various suggested methods from Google, there is either a much smaller amount, or none at all.

top indicates that there is 1.8GB of swap memory:

top - 15:35:21 up 6 days,  3:23,  1 user,  load average: 1.60, 1.43, 1.37
Tasks:  47 total,   2 running,  45 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0%us,  1.3%sy,  0.0%ni, 14.7%id, 83.8%wa,  0.0%hi,  0.0%si,  0.1%st
Mem:   3928924k total,  2855572k used,  1073352k free,   640564k buffers
Swap:        0k total,        0k used,        0k free,  1887788k cached

swapon -s doesn't seem to think there's any swap space:

Filename                                Type            Size    Used    Priority

free -m doesn't think there's any swap either:

             total       used       free     shared    buffers     cached
Mem:          3836       3663        172          0        626       2701
-/+ buffers/cache:        336       3500
Swap:            0          0          0

And neither does vmstat:

procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
 0  3      0  66224 641372 2874744    0    0    21  5012   21   33  2  2 76 19

But cat /etc/fstab thinks there is a swap partition:

/dev/xvda1 /     ext3    defaults 1 1
/dev/xvda2 /mnt  ext3    defaults 0 0
/dev/xvda3 swap  swap    defaults 0 0
none       /proc proc    defaults 0 0
none       /sys  sysfs   defaults 0 0

However df -k gives no indication of the xvda3 partition:

Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/xvda1            16513960  15675324         0 100% /
tmpfs                  1964460         8   1964452   1% /lib/init/rw
udev                   1914148        28   1914120   1% /dev
tmpfs                  1964460         4   1964456   1% /dev/shm

So I really don't know what to make of this, because I appear to have a process using about 10 times more virtual memory than what might be available, and I have no idea where this virtual memory is on the system.

I'm probably misinterpreting the output of the tools, so I'd be grateful if someone would be able to set me straight: What have I got wrong, what's the right interpretation, and how do you reach that interpretation?

EDIT0: We use 10gen's MMS for monitoring the database, the relevant section for memory from the last data point is:

 "mem": {
        "virtual": 20749,
        "bits": 64,
        "supported": true,
        "mappedWithJournal": 20376,
        "mapped": 10188,
        "resident": 1219
    },

This JSON is specific to the database process (I believe) rather than the system as a whole.

fdisk -l /dev/xvda outputs... nothing? I tried each of the 3 xvda entries in /etc/fstab as well:

root@ip:~# fdisk -l /dev/xvda1

Disk /dev/xvda1: 34.4 GB, 34359738368 bytes
255 heads, 63 sectors/track, 4177 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000

Disk /dev/xvda1 doesn't contain a valid partition table
root@ip:~# fdisk -l /dev/xvda2
root@ip:~# fdisk -l /dev/xvda3
root@ip:~#

Edit1:

Output of cat /proc/meminfo for the sake of completeness:

MemTotal:        3928924 kB
MemFree:          726600 kB
Buffers:          648368 kB
Cached:          2216556 kB
SwapCached:            0 kB
Active:          1945100 kB
Inactive:         994016 kB
Active(anon):      60476 kB
Inactive(anon):    12952 kB
Active(file):    1884624 kB
Inactive(file):   981064 kB
Unevictable:           0 kB
Mlocked:               0 kB
SwapTotal:             0 kB
SwapFree:              0 kB
Dirty:            387180 kB
Writeback:             0 kB
AnonPages:         73380 kB
Mapped:          1188260 kB
Shmem:                48 kB
Slab:             149768 kB
SReclaimable:     146076 kB
SUnreclaim:         3692 kB
KernelStack:        1104 kB
PageTables:        16096 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:     1964460 kB
Committed_AS:     305572 kB
VmallocTotal:   34359738367 kB
VmallocUsed:       16760 kB
VmallocChunk:   34359721448 kB
HardwareCorrupted:     0 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
DirectMap4k:     3932160 kB
DirectMap2M:           0 kB
share|improve this question
    
You haven't shown any output from your monitoring system. Why not? –  Michael Hampton Nov 18 '12 at 15:57
2  
Swap: 0k total, 0k used, 0k free, 1887788k cached in top says you have 0 swap, the 1.8G cached is your RAM which top displays badly. Can you post a fdisk -l /dev/xvda? –  zhenech Nov 18 '12 at 16:01
    
Thanks for the comments, I've added the extra information you (both) mentioned –  chrisbunney Nov 18 '12 at 16:13
1  
Keep in mind that MongoDB uses memory mapped files. It derives its virtual memory from these files, therefore, and not from swap space. The Mongo documentation mentions this in several places, but Caching and Checking Server Memory Usage are worth a read. –  cyberx86 Nov 19 '12 at 0:15
    
Thanks @cyberx86, I had read that, but wasn't grokking it, the hangover I had from Saturday night probably wasn't helping. Being new to the concept of memory mapped files, I wasn't sure where these files were meant to be, my reasoning being if they're virtual memory, then surely these files should be in swap space, but clearly they're not and I can see that now –  chrisbunney Nov 19 '12 at 10:12

2 Answers 2

up vote 2 down vote accepted

df will only show you mounted filesystems, not swap space. swapon -s or /proc/swaps is definitive. Your system has no swap space.

It looks like top is somewhat deceptive, since it puts the buffers value on the same line as the memory stats and the cached value on the same line as the swap stats. buffer and cache both refer to memory used by the storage subsystem and don't have anything to do with swap. (See this answer.) /proc/meminfo is a little friendlier for looking at memory stats.

share|improve this answer

Try mkswap /dev/xvda3 and then swapon -a. But only if you're absolutely sure there's nothing of value on /dev/xvda3.

share|improve this answer
    
It would appear that /dev/xvda3 doesn't exist .. I think that there's something messed up with the partitions, I've also noticed that df thinks there's 16GB of storage space when it's a 32GB disk... –  chrisbunney Nov 18 '12 at 17:14

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.