Do you wonder if your IT business is truly profitable or if you should raise your prices? Learn how to calculate your overhead burden using our free interactive tool and use it to determine the right price for your IT services. Start calculating Now!
Noob: I wish to reallocate resources for my guest servers -- still clueless how to measure what's available.
I had this question after viewing VMWare Resource Limiting.
Disclaimer: I am a senior DBA, new to VMware (and several other SA responsibilities). What I've studied so far isn't focused enough on my immediate need.
I've inherited a single server, ESXi 5.10, with several guest VMs. Each guest runs Oracle systems on RHEL. All but one of the VMs are production. I'm tasked with converting the one non-production VM into a scratch/ lab platform. This repurposed VM, which I'll call LAB1, will be staged with current versions of RHEL and Oracle,
To illustrate, say that I have three prod VMs plus LAB1. The resource distribution seen in vClient is 30-30-30-10. My target is 25-25-25-25. As far as I can tell, prod CPU usage is very low, and memory use is mostly low to moderate. I'm not aware of any resource pools, reserves, etc. It stands to reason that my alterations will come in gradual steps.
So, please help me keep from crashing production while I learn from my mistakes :) TIA.
Disclaimer: I am a senior DBA, new to VMware (and several other SA responsibilities). What I've studied so far isn't focused enough on my immediate need.
I've inherited a single server, ESXi 5.10, with several guest VMs. Each guest runs Oracle systems on RHEL. All but one of the VMs are production. I'm tasked with converting the one non-production VM into a scratch/ lab platform. This repurposed VM, which I'll call LAB1, will be staged with current versions of RHEL and Oracle,
To illustrate, say that I have three prod VMs plus LAB1. The resource distribution seen in vClient is 30-30-30-10. My target is 25-25-25-25. As far as I can tell, prod CPU usage is very low, and memory use is mostly low to moderate. I'm not aware of any resource pools, reserves, etc. It stands to reason that my alterations will come in gradual steps.
So, please help me keep from crashing production while I learn from my mistakes :) TIA.
Do more with
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
I'd take a look here.
https://labs.vmware.com/vmtj/vmware-distributed-resource-management-design-implementation-and-lessons-learned
Seems like there may be less of an issue with this due to the way vmware prioritizes resources. I have a 65 guest VM cluster and we use the default specifications from VMWare without an issue. The cluster load is about 40% during high use production.
https://labs.vmware.com/vmtj/vmware-distributed-resource-management-design-implementation-and-lessons-learned
Seems like there may be less of an issue with this due to the way vmware prioritizes resources. I have a 65 guest VM cluster and we use the default specifications from VMWare without an issue. The cluster load is about 40% during high use production.
Memory is often the bottleneck, and can be changed, and reduced without any issues.
Active Memory, is the indicator you need to look at.
Active Memory, is the indicator you need to look at.
So to prevent one machine from hogging resources would you just change the available memory on the non production server to a smaller amount? Just curious on this issue as well.
@Brian, not so. I need to increase resources to the non-production, LAB1, to become at least equal to the others -- and without crashing the production VMs as their resources shrink.
Once configured with RHEL 7 and Oracle 12c (and backed up), LAB1 will be loaded with a copy of production data, then used for user-acceptance testing prior to upgrading each production system. Perhaps I should clarify that both RHEL, and Oracle, strongly recommend a clean install of the new product, rather than a migration.
Once configured with RHEL 7 and Oracle 12c (and backed up), LAB1 will be loaded with a copy of production data, then used for user-acceptance testing prior to upgrading each production system. Perhaps I should clarify that both RHEL, and Oracle, strongly recommend a clean install of the new product, rather than a migration.
@Andrew, I'll check that out. It's pretty clear, however, that the memory load as a point-in-time doesn't tell me a thing about load over time -- say for a 24x7 week. With a historical record, I might find that PROD2 never goes over 10% CPU. I could safely pull from that one.
The system log shows that each prod server will throw a resource warning when the daily backup kicks off. I don't need to hear from my tech lead that users called her with after-the-fact complaints :)
The system log shows that each prod server will throw a resource warning when the daily backup kicks off. I don't need to hear from my tech lead that users called her with after-the-fact complaints :)
@Andrew, thank you. The article is way over my little Pooh head for now, but it will be fun to understand -- eventually. You got me looking for active memory, and this one link caught my eye: https://communities.vmware.com/docs/DOC-10398.pdf.
Think sar or top, piped to a text output.
Think sar or top, piped to a text output.
If you look at your resources for each VM, you should be able to find a value for Active Memory.
As a DBA do you not have a feel for how much memory your servers/application require ?
Issue with some applications, such as Databases, is they Grab ALL the memory, and release back to the OS, when applications require it, so therefore it appears that the Database is Actively using all memory from the start which is not helpful.
As a DBA do you not have a feel for how much memory your servers/application require ?
Issue with some applications, such as Databases, is they Grab ALL the memory, and release back to the OS, when applications require it, so therefore it appears that the Database is Actively using all memory from the start which is not helpful.
@Andrew, on current Oracle systems the memory allocation is dynamic up to pre-set limits (targets). Swapping is discouraged -- whereas what I read on this topic, vSphere appears to be just the opposite. I suspect I need to talk to VMware about what one learns in which classes.
However, the answers here have been mostly helpful, thanks. There's one follow-up question about the active memory suggestions. In the pop-up window titled Customize Performance Chart, I figured out to create a custom chart. My earlier frustration was that this tool is locked down to display only the last one hour by default. The last and the from-to controls are not available to me as root. I'd appreciate any last advice before I have to open a formal service ticket.
However, the answers here have been mostly helpful, thanks. There's one follow-up question about the active memory suggestions. In the pop-up window titled Customize Performance Chart, I figured out to create a custom chart. My earlier frustration was that this tool is locked down to display only the last one hour by default. The last and the from-to controls are not available to me as root. I'd appreciate any last advice before I have to open a formal service ticket.
A virtual server hosted on VMware vSphere is no different to a physical server. So not quite sure about your reference. Swapping only occurs on a Windows VM, if you run out of memory.
From your last post mentioning 1 hour, and root, suggests you maybe not logged into vCenter Server, which allows better reporting.
Is this just a single standalone ESXi server, and you are logged in as root ? no vCenter Server for management, and maybe you are using the FREE ESXi version ? e.g. no paid license to VMware.
From your last post mentioning 1 hour, and root, suggests you maybe not logged into vCenter Server, which allows better reporting.
Is this just a single standalone ESXi server, and you are logged in as root ? no vCenter Server for management, and maybe you are using the FREE ESXi version ? e.g. no paid license to VMware.
@Andrew, thanks -- part of my fog may be that I'm not seeing the same features that you experienced people are -- but I'm coming along. Also, I had time today to focus in on the CPU/memory/storage consumption. I finally noticed that my time window choices were still locked down, allowing me only the recent-sixty-minutes display. I did experiment with choosing the advanced reporting control. Unfortunately, that didn't seem to unlock my duration radio buttons.
My license is for Essentials, not the plus version, however.
My license is for Essentials, not the plus version, however.
Okay, you have Essentials.
Have you installed vCenter Server ?
It's also included in your license ?
vCenter Server will allow you to store the performance data, and review it, and therefore look at more than 1 hours worth.
What version of ESXi do you have ?
Have you installed vCenter Server ?
It's also included in your license ?
vCenter Server will allow you to store the performance data, and review it, and therefore look at more than 1 hours worth.
What version of ESXi do you have ?
5.10. No to the vCenter, yet. :)
It's the only choice that's presented after O/S login. I read something this week that inferred there's a submenu to provide a limited set of O/S commands, I suspect that's available in some other VM product. Further, whoever set up this master server apparently disabled ssh access.
So, my current action includes a Monday SR with VMware about the apparent locked down options (vClient prohibits changes to the one hour metrics); and to RHEL to work around the command line access. I have a weekend maintenance window coming up, my first opportunity to bounce the master server. And no, acquiring another ESXi host is not likely to happen :(
In the meantime, I've begun manually recording the VMs' consumed resources during the peak hours.
And finally, I feel like my original question has been addressed, how/where to measure the load. I'll create new questions, I'm sure, after the server bounce. I'll wait for your reply to this, then award points / close.
Regards,
dvz
It's the only choice that's presented after O/S login. I read something this week that inferred there's a submenu to provide a limited set of O/S commands, I suspect that's available in some other VM product. Further, whoever set up this master server apparently disabled ssh access.
So, my current action includes a Monday SR with VMware about the apparent locked down options (vClient prohibits changes to the one hour metrics); and to RHEL to work around the command line access. I have a weekend maintenance window coming up, my first opportunity to bounce the master server. And no, acquiring another ESXi host is not likely to happen :(
In the meantime, I've begun manually recording the VMs' consumed resources during the peak hours.
And finally, I feel like my original question has been addressed, how/where to measure the load. I'll create new questions, I'm sure, after the server bounce. I'll wait for your reply to this, then award points / close.
Regards,
dvz
5.10. No to the vCenter, yet. :)
So this is the reason why you CANNOT look at performance data, longer than you can because there is no database to store it.
For long term performance monitoring, you really should manage your server using vCenter Server, especially also as you have paid for it!
So, my current action includes a Monday SR with VMware about the apparent locked down options (vClient prohibits changes to the one hour metrics); and to RHEL to work around the command line access. I have a weekend maintenance window coming up, my first opportunity to bounce the master server. And no, acquiring another ESXi host is not likely to happen :(
Install vCenter Server..... you've not install the FULL Product!
Premium Content
You need an Expert Office subscription to comment.Start Free Trial