Yes, @PaulDND is correct. You may simply run different workloads on each IOAnalyzer worker to synthesize some workload mixes.
↧
Re: Use I/O analyzer on multiple VMs on same host ?
↧
Re: Results show ESXtop CPU Util stats only
I Had the same issue with the IO/Analyzer on the VDS switch moved it to an ESX host with VSS and everything worked as expected. Not sure why the appliance does not function well on VDS switches could be something I need to change in my config.
↧
↧
Re: Results show ESXtop CPU Util stats only
We do test both VDS and VSS in our dev environment without problems. Could you please go to /var/www/expts/ on the controller VM and then cd to the test that you think it should contain desired stats? In the directory, please then go to "timeseries" folder and do an ls to see whether IOAnalyzer did collect all disk and networking performance objects (you should see a number of timeseries-network/adapter/vm/device-* files). If under the timeseries folder, there are reasonable amount of timeseries files, most likely IOAnalyzer has filtered out non-active counters. Please try to push more traffic and re-run the test again. If problem still persists, please send an email to io-analyzer-info@vmware.com so that we could provide a mean to adjust filtering threshold. Having said that, if majority of timeseris data files are missing, IOAnalyzer might have failed to pull most of the counters from ESX. If this is the case, please send us your /var/log/daemon.log and /var/www/ioa.log so that we can collect debugging information.
Thanks,
Chien-Chia
↧
Re: I/O Analyzer 1.5 - Tests Not Running
I figured out the problem on my install. For some reason, when you select Run Now, it actually schedules the test to run about 7 minutes in the future. The easiest way to make it actually run now is save the test as something (I used test1), then go to the scheduler, select the config name, select get current server time, select schedule test and it will run.
↧
Re: I/O Analyzer 1.5 - Tests Not Running
I have the same issue running the new version on a customers site. I'll send the logs to the email address in this post. I've tryed lots of things (DNS names, ip addresses, disabled lockdown mode, started esxi shell and ssh, etc etc) still gets stuck on that dreaded "initializing VI SDK". May have to go back to the old way (io-meter).
↧
↧
Re: I/O Analyzer 1.5 - Tests Not Running
Hi Robert,
I think your case is a known bug to us and we have a quick patch for you. Please send us an email (io-analyzer-info@vmware.com) and we'll provide you the patch along with the instructions on how to apply it. We're working on releasing a minor update that includes the patch, but it'll be good if you could give it a try and see whether that fixes your issue. Sorry for the inconvenience.
Thanks,
Chien-Chia
↧
I/O Analyzer 1.5.1 has officially released.
Dear Community,
We've just released I/O Analyzer 1.5.1, which includes a detailed User's Guide and most importantly patches to quite a few bugs that have been reported so far. For those who are having "test no running" issues or "scheduler is stuck in initialization state," please download and deploy the new version. Thanks those who have reported bugs and kindly provided us log files for debugging. Please let us know if you run into issues with the new release.
Thanks,
Chien-Chia on behalf of the I/O Analyzer development team
↧
Multiple Hosts, Multiple Datastores
Hello,
I've got 3 hosts all connected to 3 shared 2TB datastores via iSCSI. If I want to verify the performance of each datastore, as well as the storage performance of each host, would it make sense to just have each host talk to one datastore (since they share paths) or should I test with each host hitting each datastore? I would think having multiple hosts talk to multiple datastores would make it easier to isolate host specific issues but I imagine most issues exist at a higher level than specific host issues (inadequate or misconfigured hardware vs. a bad NIC or something).
Thanks.
↧
Re: Multiple Hosts, Multiple Datastores
Testing iSCSI storage could be a bit tricker than FC SAN storage because users often fell into network configuration problem.
I think you can try both 1 host-1 datastore and 3 host-3 datastores tests since iSCSI datastores are already shared across hosts.
For maximum IOPS tests, create 9 VMDKs (3 VMDK for each datastores) and attach each VMDK from each datastore for each VM on each host.
In short, since iSCSI storage are shared across hosts, either way should be fine.
↧
↧
I/O analyzer results translation help
hello guys,
im new to this and just purchased a raid 6405 card (4 SATA WD RED 2TB Drives) in raid 5 config with 256k stripe with esxi 5.1 free edition. Is there a primer sheet where i can get what are acceptable results for my config or a general know how doc? i ran the I/O analzyer 1.5.1 on the vm where my raid card is installed and got the results below. to me the WRITES/s seem very low or am i just misinterpretting it wrong?
↧
Re: I/O analyzer results translation help
At the top of that screen shot you can see that you selected to run a 100% read workload, so a very small number of writes (basically enough to record the data) is expected.
↧
Re: I/O analyzer results translation help
thanks for the reply. How do my READS look based on my raid controller and SATA drives? how about a guide that explains the columns?
↧
Re: I/O analyzer results translation help
It looks like your reads are being satisfied entire by the cache on the card, because 4 SATA drives can't push 38K IO/s (generally assum around 70-90 IO/s for SATA drives).
You need a workload thats more realistic than 512byte 100% read.
Try something like 16K 50% read 50% write to get a better feeling.
↧
↧
Re: I/O analyzer results translation help
on 1.5 i only see read tests.
↧
Re: I/O analyzer results translation help
The one you have selected, OLTP 8k, is a mixed workload, for example.
↧
Re: I/O analyzer results translation help
ok ran that test for 6 min and these are the results. numbers seem very low to me. Can you let me know?
Thanks
↧
Re: I/O analyzer results translation help
Thats closer to what I'd expect. 250 IO/s for a RAID5 set of low performance SATA drives without a write cache is pretty normal. You are being mostly limited by your response times (DAVG/cmd), but again, those are on par for the low end drives in use.
Edit: Generally, we expect 7200 RPM drives to do about 70 IO/s for normal workloads. You have 4 of them, doing 70% reads (good for R5 sets). So 70*4 = 280 IO/s maximum (assuming you only did read, which you aren't). Obviously thats pretty close to what you are gettiing. If this were my array, I'd be happy that I'm hitting target performance.
↧
↧
Re: I/O analyzer results translation help
sorry meant to say raid 5 (3 disks) and 1 hot spare. I was worried about the MBPS Read and Write results which seem very low to me. thanks for all the help, I really appreciate it. Is there a white paper that can give me info about the results?
Thanks
↧
Re: I/O analyzer results translation help
Then I'd call those results very good.
The MBPS rates are low because you are doing small block (8k) IO...its expected.
I have no white paper.
↧
Re: I/O analyzer results translation help
thanks for all the help.
↧