Table of contents
Preamble
Linux memory usage: how to know how many free memory remain on your Linux server ? With the Linux default caching mechanism it might not be so easy to answer this question… Let see with few examples how to sort this out.
/dev/shm device
/dev/shm is a mount point that is created by default on most Linux operating system when you install a new server.
/dev/shm is the Linux implementation of shared memory concept. Shared because multiple programs will access the same memory portion and so sharing at high speed information (much faster than file on standard disks).
shm / shmfs is also known as tmpfs and so is a temporary file storage and even if it appears as a mount point it uses (virtual) memory and not your disk device.
Example of a 4GB /dev/shm:
[root@server1 ~]# df |grep shm tmpfs 4089276 0 4089276 0% /dev/shm [root@server1 ~]# grep shm /etc/fstab tmpfs /dev/shm tmpfs defaults 0 0 |
By default /dev/shm is created to use half size of your server physical memory:
[root@server1 ~]# free -m total used free shared buffers cached Mem: 7986 7938 48 0 61 2155 -/+ buffers/cache: 5720 2266 Swap: 5012 362 4649 |
To know if it is in use you can simply list its contain to see if files have been created. So far the only application I have seen using it is Oracle 11gR2 when using MEMORY_MAX_SIZE and MEMORY_TARGET initialization parameters (maybe SAP is also using it, to be verified).
You can obviously change its size with:
[root@server1 ~]# grep shm /etc/fstab tmpfs /dev/shm tmpfs defaults,size=2g 0 0 [root@server1 ~]# mount -o remount /dev/shm |
You can also dynamically use something like:
[root@server1 ~]# mount -o remount,size=2G /dev/shm |
This filesystem is also an open door for hackers attack that’s why many articles on Internet strongly suggest to secure it…
Remark:
For the ones wondering why not using ramfs instead (not swappable) then answer is clear in My Oracle Support (MOS) note 749851.1: not supported by Oracle and Automatic Memory Management (AMM).
I have been dubious on this /dev/shm for a long time and have always considered than the size specified when mounting it was pre-allocated and reserved for further usage: this is not true. Size is effectively allocated only when you really need it (starting an Oracle instance with AMM for example). So in clear no harm to set it half of server memory.
Free command
From Redhat support:
The philosophy in Linux is that an unused resource is a wasted resource. The kernel therefore will use as much RAM as it can to cache information from your local and remote filesystems and disks. This builds up over time as reads and writes are done on the system trying to keep the data stored in RAM as relevant as possible to the processes that have been running on your system. This caching is reported by the system as the sum of two numbers, buffers and pagecache. The cache is reclaimed, not at the time of process exit (you might start up another process soon that needs the same data), but upon demand – i.e. When you start a process that needs a lot of memory to run, the Linux kernel will reclaim memory that had been storing cached data and give it to the new process.
There are some things which get reported as cache which are not directly freeable by the kernel, such as anonymous mmaps and shm regions. These will however, report against all processes attached to them unlike normal cache which is not part of the address space of any running process but is simply a kernel mapping.
For example (Units are in megabytes):
# free -m total used free shared buffers cached Mem: 1000 900 100 0 350 350 -/+ buffers/cache: 200 800 |
In this example, as far as applications are concerned the system is using only 200MB of memory and has 800MB free and available for use if needed:
- Total Physical Memory = 1000 MB
- Physically Used Memory = 900 MB
- Actual used memory = 200 MB
- buffers = 350 MB
- cached = 350 MB
- Physically Free Memory = 100 MB
- Memory free for Applications = 800MB
Mathematical Calculations:
- Physically Used Memory = Actual used memory + buffers + cache = 200 + 350 + 350 = 900 MB
- Physically Free Memory = Total Physical Memory – Actual used memory – buffers – cache = 1000 – 200 – 350 – 350 = 100 MB
- Memory free for Applications = Total Physical Memory – Actual used memory = 1000 – 200 = 800 MB
- Memory used by Applications = Physically Used Memory – buffers – cache = 900 – 350 – 350 = 200 MB
Example:
[root@server1 /]# free -mlt total used free shared buffers cached Mem: 8007 7819 187 0 232 5804 Low: 711 625 86 0 0 0 High: 7295 7194 101 0 0 0 -/+ buffers/cache: 1782 6225 Swap: 2000 0 2000 Total: 10007 7819 2187 |
/proc/meminfo file
This file provide detailed information on memory usage, what is not obvious is fact that file is automatically updated so by clubbing it with watch you create an interesting memory monitor:
[root@server1 /]# watch cat /proc/meminfo Every 2s: cat /proc/meminfo Tue Mar 8 17:57:54 2011 total: used: free: shared: buffers: cached: Mem: 8396120064 8197312512 198807552 0 244072448 6085550080 Swap: 2097405952 0 2097405952 MemTotal: 8199336 kB MemFree: 194148 kB MemShared: 0 kB Buffers: 238352 kB Cached: 5942920 kB SwapCached: 0 kB Active: 3582784 kB ActiveAnon: 1529756 kB ActiveCache: 2053028 kB Inact_dirty: 3175336 kB Inact_laundry: 801644 kB Inact_clean: 157720 kB Inact_target: 1543496 kB HighTotal: 7470644 kB HighFree: 105800 kB LowTotal: 728692 kB LowFree: 88348 kB SwapTotal: 2048248 kB SwapFree: 2048248 kB CommitLimit: 6147916 kB Committed_AS: 3295608 kB HugePages_Total: 0 HugePages_Free: 0 Hugepagesize: 2048 kB |
Free and /proc/meminfo reconciliation
[root@server1 ~]# free -m;cat /proc/meminfo total used free shared buffers cached Mem: 8007 7817 189 0 232 5804 -/+ buffers/cache: 1780 6226 Swap: 2000 0 2000 total: used: free: shared: buffers: cached: Mem: 8396120064 8197431296 198688768 0 244133888 6086209536 Swap: 2097405952 0 2097405952 MemTotal: 8199336 kB MemFree: 194032 kB MemShared: 0 kB Buffers: 238412 kB Cached: 5943564 kB SwapCached: 0 kB Active: 3581808 kB ActiveAnon: 1528720 kB ActiveCache: 2053088 kB Inact_dirty: 3175740 kB Inact_laundry: 801884 kB Inact_clean: 157720 kB Inact_target: 1543428 kB HighTotal: 7470644 kB HighFree: 105744 kB LowTotal: 728692 kB LowFree: 88288 kB SwapTotal: 2048248 kB SwapFree: 2048248 kB CommitLimit: 6147916 kB Committed_AS: 3294440 kB HugePages_Total: 0 HugePages_Free: 0 Hugepagesize: 2048 kB |
MemTotal – MemFree – Buffers – Cached = 8199336 – 194032 – 238412 – 5943564 = 1823328 i.e. 1780MB so memory used of free command.
To go further
[root@server1 ~]# slabtop --sort=c Active / Total Objects (% used) : 798325 / 926847 (86.1%) Active / Total Slabs (% used) : 66730 / 69357 (96.2%) Active / Total Caches (% used) : 48 / 75 (64.0%) Active / Total Size (% used) : 231839.90K / 249148.15K (93.1%) Minimum / Average / Maximum Object : 0.01K / 0.27K / 128.00K OBJS ACTIVE USE OBJ SIZE SLABS OBJ/SLAB CACHE SIZE NAME 309911 309911 100% 0.50K 44273 8 177092K inode_cache 316860 316860 100% 0.12K 10562 32 42248K dentry_cache 82495 29323 35% 0.11K 2357 37 9428K buffer_head 1678 1558 92% 4.00K 1678 1 6712K size-4096 44460 16236 36% 0.12K 1482 32 5928K pte_chain 2098 1378 65% 2.00K 1049 2 4196K size-2048 821 821 100% 4.00K 821 1 3284K names_cache 1286 1046 81% 1.38K 643 2 2572K sock 1124 1124 100% 1.38K 562 2 2248K sighand_cache 7035 6709 95% 0.25K 469 16 1876K size-256 3672 2633 71% 0.50K 459 8 1836K size-512 25578 25518 99% 0.06K 441 64 1764K size-32 11910 8385 70% 0.12K 397 32 1588K size-128 1248 1127 90% 1.00K 312 4 1248K size-1024 3010 1671 55% 0.38K 301 10 1204K nfs_read_data 6808 2272 33% 0.16K 296 24 1184K skbuff_head_cache 2051 1821 88% 0.50K 293 8 1172K files_cache 2660 2164 81% 0.38K 266 10 1064K nfs_write_data 7710 7381 95% 0.12K 257 32 1028K size-64 6960 6390 91% 0.13K 240 31 960K sigqueue 6840 5852 85% 0.12K 228 32 912K tcp_open_request 6840 3203 46% 0.12K 228 32 912K nfs_page 2000 1798 89% 0.38K 200 10 800K mm_struct 10584 8909 84% 0.07K 189 60 756K vm_area_struct 13167 3269 24% 0.05K 171 85 684K journal_head 2355 1264 53% 0.25K 157 16 628K ip_dst_cache 3450 2048 59% 0.12K 115 32 460K blkdev_requests 4400 2610 59% 0.09K 110 42 440K file_lock_cache 3030 1875 61% 0.12K 101 32 404K tcp_tw_bucket 25 20 80% 16.00K 25 1 400K size-16384 2970 2707 91% 0.12K 99 32 396K filp 5162 3201 62% 0.06K 89 64 356K signal_cache 1275 254 19% 0.25K 85 16 340K arp_cache 4466 3509 78% 0.06K 77 64 308K cdev_cache 37 11 29% 8.00K 37 1 296K size-8192 4060 3344 82% 0.06K 70 64 280K fs_cache 3886 3322 85% 0.06K 67 64 268K pae_pgd 2 2 100% 64.00K 2 1 128K size-65536 1798 511 28% 0.06K 31 64 124K bdev_cache 1798 1263 70% 0.06K 31 64 124K inet_peer_cache 2576 1883 73% 0.03K 23 128 92K tcp_bind_bucket 2240 1468 65% 0.03K 20 128 80K revoke_record 1792 1527 85% 0.03K 16 128 64K uid_cache 1008 10 0% 0.03K 9 128 36K ip_fib_hash 1 0 0% 32.00K 1 1 32K size-32768 112 112 100% 0.24K 7 16 28K kmem_cache 348 20 5% 0.06K 6 64 24K mnt_cache 1250 5 0% 0.01K 5 341 20K revoke_table 0 0 0% 0.06K 0 64 0K size-32(DMA) 0 0 0% 0.12K 0 32 0K size-64(DMA) |
In the above inodes cache (files references, inode_cache) size 172MB and directory names (dentry_cache) size 41MB.
References
- What is /dev/shm and its practical usage
- Tips & Tricks Featured Article: /proc/meminfo Explained
- Overview of RAMFS and TMPFS on Linux
- HugePages and Oracle Database 11g Automatic Memory Management (AMM) on Linux [ID 749851.1]
- Oracle 11g internals part 1: Automatic Memory Management
vmckrish says:
Kudos!! for sharing and insightful. Yannick Jaquier