Ad Widget

Collapse

Problem with Zabbix Memory Monitoring (Linux)

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • lostmarbles
    Member
    • Mar 2009
    • 50

    #1

    Problem with Zabbix Memory Monitoring (Linux)

    We are using the Linux template to monitor Memory on a Linux box. Zabbix provides the following:
    • Buffers memory
    • Cached memory
    • Free memory
    • Shared memory
    • Total memory


    This list above is missing one of the most important Linux VM metrics, which is:
    • Inactive memory


    Without showing the Inactive Memory stat, the stats are not very useful, since most Free memory moves to Inactive memory and Linux reclaims Inactive memory when it needs to. So, the basic Linux template needs to also monitor Inactive memory.
  • richlv
    Senior Member
    Zabbix Certified Trainer
    Zabbix Certified SpecialistZabbix Certified Professional
    • Oct 2005
    • 3112

    #2
    um, what's inactive memory ? i don't remember seeing such a term.
    Zabbix 3.0 Network Monitoring book

    Comment

    • lostmarbles
      Member
      • Mar 2009
      • 50

      #3

      Google search for Inactive memory linux
      Last edited by lostmarbles; 26-03-2009, 21:10.

      Comment

      • richlv
        Senior Member
        Zabbix Certified Trainer
        Zabbix Certified SpecialistZabbix Certified Professional
        • Oct 2005
        • 3112

        #4
        well, as you didn't give specific references, i can only assume you refer to active/inactive split that is already included in other metrics.
        while i agree that monitoring active/inactive could be useful in some cases, i don't think system can _reclaim_ any of that memory, after all, it is being used.
        Zabbix 3.0 Network Monitoring book

        Comment

        • lostmarbles
          Member
          • Mar 2009
          • 50

          #5
          Originally posted by richlv
          well, as you didn't give specific references, i can only assume you refer to active/inactive split that is already included in other metrics.
          No, it is not included in other metrics. The definition of Inactive memory in a Linux system is very straight forward and must be included in the basic memory stats. It is not and it should be.

          Originally posted by richlv
          i don't think system can _reclaim_ any of that memory, after all, it is being used.
          I appreciate that you are replying to this thread to be helpful, but you simply do not have knowledge of what you are talking about, in this case. You (1) admittedly never heard of one of the most important parameters in Linux VM management (Inactive memory) and (2) seemingly are complaining that I am not teaching you about how Linux VM works!

          Your latest reply, is wrong; please be careful and research your replies before you post because others need to get correct information:

          Originally posted by richlv
          i don't think system can _reclaim_ any of that memory, after all, it is being used.
          Of course Linux reclaims Inactive memory, it is a fundamental part of how Linux management VM.

          I don't want to sound impatient or hasty, but I am a bit busy in other forums and did not post here to teach nor explain the details of how Linux manages VM - that knowledge is readily available with a bit of Google homework and document reading (if you want to know about it).

          I posted to state a fact about a missing Memory Item in the core Linux_Template. The fact is that one of the most important metrics to monitor in Linux VM is the amount of Inactive memory. Without monitoring Inactive memory, monitoring the amount of Free memory is useless. Linux can have nearly zero Free memory and 12 GB of Inactive memory and Linux will reclaim it as needed, because that is how Linux works.

          This is not a point to debate, so please don't debate it! It is an fact easily found from homework on the net and folks who understand even the basics on how Linux VM works understand this. However, it does require a bit of reading and understanding.

          I don't want to waste time arguing this, especially if you are not the person who wrote that template. Did you write it?

          Sorry, but if you don't understand the basics of Linux VM, kindly do not ask me to explain it to you; please be respectful and go read the many documents on the net about it before responding.

          It might be more useful if someone just posted who wrote that template, so I can tell them what they need to fix.

          Thanks !!

          Comment

          • Calimero
            Senior Member
            • Nov 2006
            • 481

            #6
            lostmarbles, I don't see any patch to zabbix_agent to collect the missing metric attached to your posts. Did you forget attaching the patch or are you just one of these users that prefer ranting to contributing ?

            I think it would be pretty straightforward for a highly skilled computer wizard as you are. (substract cached and buffers memory from used memory ?)

            Comment

            • jbillma
              Junior Member
              • May 2010
              • 4

              #7
              I believe this metric is equivalent to vm.memory.size[available], which appears to be the sum of free, buffered, and cached memory. I would appreciate it if someone could validate that, however, as I am not the template's author, nor did I have anything to do with developing this monitoring item. (By the way, this item is not presently in the template for Linux.)

              Comment

              Working...