Dear Lazyweb;
Wondering if there's a sysstat/nfsstat type command that will show me which "files" (or which initiating NFS clients) are absorbing or generating the most traffic?
Trying to figure out which VMDK files or ESX clients are to blame for a recent spike in activity.
Ray
On Aug 13, 2013, at 4:57 PM, Ray Van Dolson rvandolson@esri.com wrote:
Dear Lazyweb;
Wondering if there's a sysstat/nfsstat type command that will show me which "files" (or which initiating NFS clients) are absorbing or generating the most traffic?
Trying to figure out which VMDK files or ESX clients are to blame for a recent spike in activity.
There used to be an old perl script called 'netapp-top.pl' that would give you the client host and NFS I/O per host.
That script is still in the "downloads, tool chest" on the now site.
If you just want to narrow it to an ESXi host, you can use nfsstat -l (Assumes per client statistics are enabled.)
On 2013-08-13, at 4:44 PM, Mike Horwath drechsau@gmail.com wrote:
On Aug 13, 2013, at 4:57 PM, Ray Van Dolson rvandolson@esri.com wrote:
Dear Lazyweb;
Wondering if there's a sysstat/nfsstat type command that will show me which "files" (or which initiating NFS clients) are absorbing or generating the most traffic?
Trying to figure out which VMDK files or ESX clients are to blame for a recent spike in activity.
There used to be an old perl script called 'netapp-top.pl' that would give you the client host and NFS I/O per host. _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
There are several ways to do this:
Spot checks (no trending)
esxtop: v (for vscsi, r/w to sort by read/write ops)
Also Check out VisualEsxtop
http://www.virtuallyghetto.com/2013/07/how-to-run-vmwares-new-fling.html
nfsstat -l shows which (ESX) clients are using biggest % IO
Get a trial of vCenter Operations - plug it into you vcenter and within 30 minutes you can see via the Per VM IO heatmaps (over 5 minute intervals) which VMs are hogging the IO and which datastores are suffering the worst latency:
https://my.vmware.com/web/vmware/evalcenter?p=vcenter-ops57
Also keep in mind other ontap operations like snapmirror or dedup are capable of adding up to a large percentage of your aggregate's IO capacity - (50% in our case before we tuned the snapmirror schedules) Netapp NMC is a good tool for lookign at aggregate IO load
http://www.vmadmin.info/2010/07/vmware-and-netapp-deconstructing.html
Misalignment - use the netapp mbrscan tools to check for misaligned VMs, use nfsstat -d to show misaligned file level IO.
Try the vCops trial first - those heatmaps were instrumental in us migrating/isolating IO workloads. Also check NMC to see your aggregate IO load vs capacity
Happy hunting!
On Aug 13, 2013, at 2:57 PM, Ray Van Dolson rvandolson@esri.com wrote:
Dear Lazyweb;
Wondering if there's a sysstat/nfsstat type command that will show me which "files" (or which initiating NFS clients) are absorbing or generating the most traffic?
Trying to figure out which VMDK files or ESX clients are to blame for a recent spike in activity.
Ray _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
-
Colin Bieberstein -
Fletcher Cocquyt -
Mike Horwath -
Ray Van Dolson