cancel
Showing results for 
Search instead for 
Did you mean: 
Create Post

FEATURE REQUEST - Showing the correct used percent of Linux FileSystems using Net-SNMP

FEATURE REQUEST - Showing the correct used percent of Linux FileSystems using Net-SNMP

Note: These request is related to case 352423.

When monitoring Linux Filesystem using Net-SNMP with Orion, the used percent that it shows is lower than real used percent.

The situation is that GNU/Linux systems (and the like) generally have a reserved block, which is (generally) by default 5%. The reserved block is inaccessible for common use and reserved for system (for it's safety should the disk run out of space). Because the reserved block is not generally accessible it's often excluded from the reported used space. This results in slightly awkward reporting of disk space usage.

For example, this is what Orion shows: (see picture "filesystem in Orion.png" attach)

And these is what shows the Linux server using the df-k command:

Filesystem                                        1K-blocks            Used                     Available            Use%    Mounted on

/dev/mapper/vg00-usr                1523568               1114532               330400                 78%       /usr

Now the tricky part is that the standard SNMP hrStorage OID only provides Size and Usage, but not the available space. And if you don't know which percent is used as inodes space (reserved block), you can't calculate the available space.

Anyway, in a previous case that we have opened (Case 308578), SolarWinds Technical Support Team told as that for Linux using net-snmp, Orion takes some metrics using OIDs different from standards.

For example, to obtain the CPU Utilization they said the following to as (see the email attached):

For Net-SNMP, Orion uses OID= 1.3.6.1.4.1.2021.11.11.0 to search for Cpu Idle.

The OID = hrProcessorLoad  1.3.6.1.2.1.25.3.3.1.2. is for Windows and Generics.

And for FileSystems, Linux using net-snmp also has an enterprise mib that gives all the information that doesn’t show the standard as you can see in the picture attached (dskPercent.JPG)

We know that we could obtain these metrics using Universal Device Poller but that solution has two problems:
• The first and less important: We are not using APM together with NPM so we don’t have UnDP in the Orion server with APM.
• Second: We would like to have all servers monitored in the same manner. It’s more simple to administrate and more simple for users to see the information. We don’t want to have volumes of Linux servers monitored in one manner and volumes of Windows servers in another way.

We want to have an unified view and we would like to see all Volumes in the same manner, not some using the default view of Orion for Volumes and another using the custom poller view. (see picture "filesystems.jpg" attach).

According to Orion description, it supports Linux monitoring and we think that as you could obtain the CPU Idle from an enterprise mib for Linux using net-snmp, you could also obtain the FileSystems from an enterprise OID like UCD-SNMP-MIB:dskPercent (OID: 1.3.6.1.4.1.2021.9.1.9).

We think that as Orion shows CPU, Memory and Volumes in the same manner for all devices but obtains these metrics from different OIDs depending on which type of device is, it would be great if Orion can solve these problem from his side and shows as the correct information all in the same manner despite of how Orion obtains the metric.

I hope we could see these problem solved in a future release.

Best Regards,

Diego Mole

Tenaris Argentina

17 Comments
Level 14

This needs to fixed. I had lots of questions about this issue.

Level 10

Yes indeed, we need a fix for this!  We also had issues of Linux Filesystems filling up before SolarWinds would alert us due to this issue so we now use a SAM Template with a UNIX bash script monitor which uses df output instead, becasue it is much more accurate!  Now the users are a bit confused becasue the Node Details screen "Volume" view modules show different infomration than reality, which is NEVER a good thing.  Also, we require further training of users so they know to use the SAM Template info for their Graphs instead of the "Volume" information coming from SolarWinds.  Confusiton all around and must get fixed so we can relay on the "Volume" feature instead of having to use a SAM Template for filesystem disk space monitoring on Linux Servers.

Level 10

I know I'm posting to a 2yr old thread but additional info can be found here.

Re: Linux Drive Monitors not Accounting for Reserved Space

Level 8

Hi! I'm the creator of this feature request. I've been out for a long time but I would like to know if SW developers team have been working on modifying the way the filesystems information in net-snmp servers is polled by Orion.

The workarround using an script is really ugly and the users don't understand why the information is located in different places according to the type of server.

Any update??? Thanks in advance!

Level 10

dfmole There's a feature request to have a linux agent developed to address this and many other issues. I'm not sure of any other way besides the script at this time.

AGENT FOR MONITORING LINUX/UNIX OPERATING SYSTEMS

Level 8

Oh, OK! That´s also a good idea. Not as good as having the problem solved from SW side, but it's a better solution than having scripts runing in SAM module. Anyway, I insist on Solarwinds finding a definitely solution for this issue. If they can pick the information from the right place accoding to the vendor Orion sees by SNMP, the product will be much more relialable for it customers and SW will go to the top!

Somewhat unrelated chaps, but I came across this thread the other day:

Poller for "LinuxAteMyRam" now available!

It certainly helped some of my customers sort out memory usage reported!

Level 9

Did anyone on this thread implement linuxatemyram successfully?

I implemented it, and have no data to show on my node screens. The data is poll-able (tested before deployed), but now, the cpu and memory graph is completely empty, and data ceased to be collected since the date/time I switched to using linuxatemyram.

I left linuxatemyram untouched (i.e., did not modify anything), and tested against a net-snmp Linux node before deploying. All data returned was valid.

Here are some pics. Hopefully you will see something that I'm completely missing?:

2015-12-02 13_08_59-Create Poller Package.png

I made a copy of the running poller so that I could test it and see whether the data was being returned, which the image above shows it indeed is.

And here you see that it is applied to the node:

2015-12-02 13_06_10-List Resources.png

2015-12-02 12_57_21-Node Details - Summary .png2015-12-02 12_58_14-Custom Chart.png

Here you can see that I had good cpu data through till I deployed linuxatemyram

2015-12-02 12_59_46-Node Details - Summary - MinMaxAverage of CPU Load.png

2015-12-02 13_01_30-Custom Chart - MinMaxAvg for November.png

Product Manager
Product Manager

SAM 6.3 Beta 2 is now available which includes a Linux Agent for Node, Volume, Interface, and Application monitoring. This agent should address many of the shortcomings associated with monitoring Linux host via SNMP, up to and including properly calculating volume usage statistics referenced in this thread. If you already own Server & Application Monitor and are under active maintenance, you can sign-up to participate in the beta at the link below.

Hold the phone! That's awesome news!