Redhat Memory Question
list Wes Neal
I have never monitored any Redhat systems before, all Solaris in the past for the most part. Since top and free always show all the physical memory in use on Redhat, how do people monitor memory usage accurately with Xymon? Currently the Memory tag shows as yellow under this new Redhat server, though in my mind it should be green.
Mem: 16435916k total, 16311612k used, 124304k free, 6255112k buffers
Swap: 18481144k total, 0k used, 18481144k free, 6391000k cached
total used free shared buffers cached
Mem: 16435916 16311588 124328 0 6255116 6391040
-/+ buffers/cache: 3665432 12770484
Swap: 18481144 0 18481144
Thanks in advance
Wes Neal
list Mark Felder
Honestly, memory usage should never go red unless there is enough memory pressure to generate a lot of swap paging I/O. Maximizing memory usage is a feature of a modern OS. I'm surprised your swap is completely empty on that server. Swap is meant to be used.
list Jeremy Laidman
On Fri, Mar 20, 2015 at 2:25 AM Neal, Jonathan W via Xymon <xymon at xymon.com>
▸
wrote:
I have never monitored any Redhat systems before, all Solaris in the past for the most part. Since top and free always show all the physical memory in use on Redhat, how do people monitor memory usage accurately with Xymon?
Xymon tracks swap usage, physical memory usage, and "actual" memory usage.
The last of these is what most people care about. In analysis.cfg you can
set thresholds on any of these. On my system the thresholds for physical
memory usage are 100% for yellow and 101% for red.
Under Linux, "actual memory" is obtained from the output of "free" which is
recorded in the [free] section of the client data, specifically from the
"buffers/cache" line. The value for "actual used" comes from the first
number. The "actual total" is the total physical memory and is the first
number in the "Mem:" line.
▸
Currently the Memory tag shows as yellow under this new Redhat server,though in my mind it should be green.
For actual memory usage, yes.
▸
Mem: 16435916k total, 16311612k used, 124304k free, 6255112k buffersSwap: 18481144k total, 0k used, 18481144k free, 6391000k cached
total used free shared buffers cached
Mem: 16435916 16311588 124328 0 6255116 6391040
-/+ buffers/cache: 3665432 12770484
Swap: 18481144 0 18481144
In your case, on your "memory" status page you would see "Actual 3579M 16050M 22%" meaning 3.5GB non-cache/buffer RAM usage used out of 16GB total physical RAM, which is 22%. So actual usage is 22%. If your yellow threshold for MEMACT is the default of 90, then this 22% is well short of triggering a yellow condition. And it must be something else. On your memory status page each line should have its own coloured dot. Which of the lines has a yellow dot? I'm guessing it's the "Physical", and someone has modified the MEMPHYS threshold in analysis.cfg to go yellow at 99% instead of 100%. J
list Paul Root
I have a script that frees up those buffers that make memory look full
sync
echo 3 >/proc/sys/vm/drop_caches
free -mt
▸
-----Original Message-----
From: Neal, Jonathan W [mailto:user-9e31f92d698c@xymon.invalid]
Sent: Thursday, March 19, 2015 10:15 AM
To: Xymon MailingList
Subject: Redhat Memory Question
I have never monitored any Redhat systems before, all Solaris in the past for the most part. Since top and free always show all the physical memory in use on Redhat, how do people monitor memory usage accurately with Xymon? Currently the Memory tag shows as yellow under this new Redhat server, though in my mind it should be green.
Mem: 16435916k total, 16311612k used, 124304k free, 6255112k buffers
Swap: 18481144k total, 0k used, 18481144k free, 6391000k cached
total used free shared buffers cached
Mem: 16435916 16311588 124328 0 6255116 6391040
-/+ buffers/cache: 3665432 12770484
Swap: 18481144 0 18481144
Thanks in advance
Wes Neal
This communication is the property of CenturyLink and may contain confidential or privileged information. Unauthorized use of this communication is strictly prohibited and may be unlawful. If you have received this communication in error, please immediately notify the sender by reply e-mail and destroy all copies of the communication and any attachments.
list Japheth Cleaver
▸
On Thu, March 19, 2015 5:19 pm, Jeremy Laidman wrote:
On Fri, Mar 20, 2015 at 2:25 AM Neal, Jonathan W via Xymon <xymon at xymon.com> wrote:I have never monitored any Redhat systems before, all Solaris in the past for the most part. Since top and free always show all the physical memory in use on Redhat, how do people monitor memory usage accurately with Xymon?Xymon tracks swap usage, physical memory usage, and "actual" memory usage. The last of these is what most people care about. In analysis.cfg you can set thresholds on any of these. On my system the thresholds for physical memory usage are 100% for yellow and 101% for red. Under Linux, "actual memory" is obtained from the output of "free" which is recorded in the [free] section of the client data, specifically from the "buffers/cache" line. The value for "actual used" comes from the first number. The "actual total" is the total physical memory and is the first number in the "Mem:" line. Currently the Memory tag shows as yellow under this new Redhat server,though in my mind it should be green.For actual memory usage, yes. Mem: 16435916k total, 16311612k used, 124304k free, 6255112k buffersSwap: 18481144k total, 0k used, 18481144k free, 6391000k cached total used free shared buffers cached Mem: 16435916 16311588 124328 0 6255116 6391040 -/+ buffers/cache: 3665432 12770484 Swap: 18481144 0 18481144In your case, on your "memory" status page you would see "Actual 3579M 16050M 22%" meaning 3.5GB non-cache/buffer RAM usage used out of 16GB total physical RAM, which is 22%. So actual usage is 22%. If your yellow threshold for MEMACT is the default of 90, then this 22% is well short of triggering a yellow condition. And it must be something else. On your memory status page each line should have its own coloured dot. Which of the lines has a yellow dot? I'm guessing it's the "Physical", and someone has modified the MEMPHYS threshold in analysis.cfg to go yellow at 99% instead of 100%. J
The whole concept of 'free memory' is really a bit tricky. Linux 3.14+ has provided a solution in /proc/meminfo (https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=34e431b0ae398fc54ea69ff85ec700722c9da773), but it's not available everywhere. Furthermore, the 'free -a' output depends on your specific version of procps(-ng) and possibly your sysctl setting of meminfo_legacy_layout (https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html-single/6.6_Technical_Notes/index.html#ch-Important-Changes-to-External-Kernel-Parameters). Regards, -jc
list Mark Felder
▸
On Thu, Mar 19, 2015, at 19:35, Root, Paul T wrote:
I have a script that frees up those buffers that make memory look full
sync
echo 3 >/proc/sys/vm/drop_caches
free -mt
Why would you do that? It is strongly recommended to not do that:
Use of this file can cause performance problems. Since it discards cached objects, it may cost a significant amount of I/O and CPU to recreate the dropped objects, especially if they were under heavy use. Because of this, use outside of a testing or debugging environment is not recommended.
Page cache is not memory that is unavailable to other applications. If a program requests more memory and there is not enough free, it will take the page cache.