There's already a Windows Agent in the works for SAM that addresses most of the limitations associated with Agentless monitoring of Windows hosts. What about an Agent that addressed the same/similar limitations for Linux/Unix operating systems?
Yes...we need it. (linux, AIX, HP/UX, possibly solaris)
In for a penny, in for a pound. If you guys do it for Winders then you certainly need to do it for systems that you have far less visibility into such as Linux/Unix etc.
I'm deeply DEEPLY conflicted about agent-based monitoring. On the one hand, it bypasses a lot of the issues around security, polling frequency, etc.
On the other, in every implementation of agent-based monitoring I've seen (two different Tivoli installs, BMC Patrol, NetIQ), the agent-to-staff ratio has been about 1,000 to 1.
So I'm worried about what this will mean for skin-and-bones monitoring teams that only survive because agentless has such dramatically lower staff costs.
That said, IF you have agents for Windows, you absolutely need to have them for Linux (Centos/RedHat and SUSE if you need specifics) along with AIX, and HPUX. Solaris and Tru64 are nice-to-have's after that.
I have lived that world...in a large environment you need to monitor and automate as close to the source as possible otherwise your management platform is having to do all the work. You need to switch to a management by exception model. Only send the exceptions up to reduce all the noise and traffic.
Leon Adato; I would certainly agree with your concerns as I have had those problems problems as well. My expectation is that using the agent will be a new option and not the only way to do things. If that is the case you can pick and choose to use it only when it makes sense.
An agent is a quick fix for companies with the Network ACL rule of deny any any as the standard! My company chose SolarWinds as our enterprise standard but as we rolled it out to our environment we were plagued by some business lines where the IT folks do things a little differently. My company acquires companies now and then and the systems are rarely integrated without a ton of pain and drama. This would allow us to roll out the standardized monitoring solution (SolarWinds and all of the integrated apps) on day 15 (since we have a 2 week lead time on a Network ACL implementation) until the full ACL for agentless monitoring can have all the ports agreed upon.
Having the agent option gives you tools that you may not otherwise have. I'd rather have the option available to use than not have it.
I know here I would use it without hesitation.
An agent for UNIX/Linux would be amazing. ONE BIG request would be to differentiate between the actual disk usage and the reserved space. Another request is to have it so the agent does not need to run as root.
Linux Drive Monitors not Accounting for Reserved Space
I have been posting to Support and Sales about this ago.
The biggest issue for our company when using your products were:
2) Fake alerts due to net-snmp on Linux not sending complete process or other tables resulting in fake alerts (i.e. imagine packet loss). A Linux agent with TCP communciation to the monitoring server will resolve this.
1) We need to scale quick and automate stuff including adding and removing hosts from monitoring. It would take time to automate adding/removing nodes and assigning them templates. It would have been much better if we can install an agent on every node then the node to also have a configuration which specifies which templates to be assigned to it. Then we also need one more thing - a command to remove the node from monitoring and voila - we can easily scale our monitoring to cover hosts running on the cloud behind NAT.
Of course you will also need to change a bit your pricing schema - i.e. pay by node would fit nice.
Many companies will be able to use you when on the cloud and you know a lot of companies run their software on private and public clouds now or have this intent.
I would love to see something around management of Solaris Containers. Kind of like what the Virtualization tab does for VMWare but for Sun Solaris. This is a huge issue for use as we don't have an easy way of relating containers to LDOMs to physical hardware. Being able to easily see this and to be able to set alerts based on dependencies would be great.
And AIX LPARs...
For some reason I thought it already had an agent for all major OSs. Disappointing that I did not see that as a missing key component.
AIX, HPUX and Solaris
A linux/Unix agent is a critical missing piece. It should have an option to send data to the SW server every x minutes rather than being polled. This will allow it to be used in critical networks where the Firewall prevents traffic into a client, since it is only sending data out, and no need to change any ports/ rules. Just need a single entry on the SW server to allow data in.
It is ironic that the platform that has the most issues with data quality (CPU, Volumes, Traffic all common) and therefore would benefit from an agent (I am deliberately ignoring the fact that the NetSNMP is already an agent) is the *Nix platform.
We need an agent for Solaris and Linux
Having a monitoring agent for Linux/Unix is necessary. Agent-less monitoring can only do so much.
Leon Adato, I understand your conflicting opinions about agentless vs agent based monitoring, ad share your point of view that managing agents can be more time consuming than there are hours in the day, as I also have to manage the MacAfee agents that we have at work. and yes when an agent breaks I can kill a couple of hours easily trying to get it working again, while the self healing has gotten better, agents do how ever have their users, for example agents sitting in a DMZ, allows me to only have the ports open that are required for agent to server communication, instead of having Swiss chesse for a firewall because of WMI monitoring on a windows host.
A good agent will have store and forward capability so that is a link is down, monitoring data is not lost. In the current configuration if an intermediate network device is down...all polling data for a node is lost during that period of time leaving gaps in your data....
So want this.
Using snmp to monitor processes remotely is a real hit or miss affair right now especially when there are lots of processes running
if you will implement the SAM with agent, how much of resources would it take? What if the Linux/ solaris server has been monitored through LEM wherein we also have agent to install on the end device and you would also want this server to monitor on SAM because of the APPLICATION TEMPLATE/ component monitor wizard? Does it means that I have to install the agent for SAM too? Or is the agent would be optional to install?
So, perhaps we should switch back to using our OpenView environment again? See, this is the reason we moved from OpenView to SolarWinds in the first place: so there would not be an agent to contend with and maintain. I vote No.
As with the WIndows Agent, the Linux agent would also be optional. There are certain scenarios where an agent is advantageous such as environments where overlapping IP address space is a problem. The intent would not to require an agent, but simply provide it as an option for those situations where agentless monitoring is not possible or practical.
I think the option of being able to use an agent where practical is the big win here.
Orion doesn't do unix as well as windows..and it doesn't manage things in the DMZ well either.
Of course you can go the route of a polled agent like nagios, but having an actual agent that does most of the work at the source significantly reduces polling overhead.
I am still a firm believer in doing your monitoring/automation at the source..then your primary environment is only having to take in data sent to it and processing out alerts.
There are environments where SNMP is disabled so using an agent or other methods to gather data are the only options that are available.
I worked at places where HP Openview was used and using an agent (OMx) was not mandatory and in some cases it was not practical or doable. (Sitescope, NNM, etc.)
You have the option of using SNMP or agent or even both. You can customize an agent to collect almost any kind of metrics whereby with SNMP your options are limited.
As several people have mentioned, SNMP is not the only monitoring option and in some cases an agent is your only option or is the best one.
As for maintenance (using agents) that depends on how automated your environment. We used to manually install and update Openview agents so I know how much time and trouble it can be.
It's good to have additional monitoring options. This is really needed in lots of cases. I vote Yes.
I hope this goes somewhere. It'd be nice to have a way to deal with that our nix team doesnt' want to create accounts on servers for audit reasons but is 100% okay with using an agent for permissions. Not needing ACL's = huge plus.
We need agent for HPUX
This would be great. We have a lot of linux machines over here in different flavours. On a number of them we cannot/don't have SNMP installed so this would be an outcome for us.
SAM 6.3 Beta 2 is now available which includes a Linux Agent for Node, Volume, Interface, and Application monitoring. 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.
aLTeReGo So it sounds like it pretty much does all of the same stuff the Windows agent does, is that correct?
