Azure experience

[vmoskalenko]

I have pre-production environment in Azure and we are building HA Cluster based on MS SQL Standard 2012 and StarWind v8 (build 8730)

Time-to-time I have message like below "High Availability Device "iqn.2008-08.com.starwindsoftware:NodeB-sw-data1", critical device response time has been detected. (IO operation delay is more than 10 sec). To avoid performance issues on the whole device, automatic synchronization attempts will be performed every 30 minute(s)."

I know I have slow disks
How to increase the timeout from 10 seconds up to 30 seconds?

[vmoskalenko]

And the second question, how to reduce automatic synchronization attempts delay from 30 seconds to 10 seconds ?

[darklight]

I've been suffering from such kind of warnings on my older homelab servers. After some digging I have found following parameters in c:\Program Files\StarWind Software\StarWind\StarWind.cfg:

Code: Select all

      <CmdExecTimeWarningLimitInMs value="20000"/>
      <!--If command execution time is more than this value(milliseconds) then warning notification will be generated in the events journal.
                By default, this value = 5000 milliseconds. Use 0 to disable notification.-->
      <!--<AutoSyncPeriodForPerfDegInMin value="0"/> -->
      <!--If underlying storage of HA device suffers from performance degradation then this value(minutes) defines period for automatic synchronization attempts.
                By default, this value = 30 minutes. Use 0 to disable automatic synchronization for device with performance degradation.-->

Try to adjust these ones and see if it helps.

[darklight]

The other point is that you are facing such issues in Azure. It's rather weird, so looks like something is misconfigured. Could you please give a brief description of your configuration and goals you are trying to achieve? Is it a POC or production?

[vmoskalenko]

The other point is that you are facing such issues in Azure. It's rather weird, so looks like something is misconfigured. Could you please give a brief description of your configuration and goals you are trying to achieve? Is it a POC or production?

It's pre-production environment. We enhance our existing Azure production environment into High Availability mode.
We use DS14 servers (16xCPU, 112GB RAM, Max 50K IOPS) and Premium-LRS 1Tb disks.

I test disk speed on Azure and my laptop. Please find results in the table below.

Code: Select all

--------------------------------------------------------------+---------------------------------------------------------------------------------
 Disk type                           Block Size   Write ratio | Read MB/s   Read latency   Read IOPS   Write MB/s   Write latency   Write IOPS  
--------------------------------------------------------------+---------------------------------------------------------------------------------
 Azure P30                           64K          10%         |     459.1          15.11        7345        50.89          20.848          814  
 Azure P30                           8K           10%         |    356.75          0.595       45663        39.84          19.752         5099  
 Azure Local SSD (Temporary drive)   64K          10%         |    460.12         15.398        7361        51.39           17.77          822  
 Azure Local SSD (Temporary drive)   8K           10%         |    459.17          2.003       58773        51.22           1.548         6555  
 Laptop SATA                         64K          10%         |     20.89         353.19         334         2.28           358.5           36  
 Laptop SATA                         8K           10%         |      4.41         203.59         563         0.49           210.7           62  
 Laptop SSD                          64K          10%         |    238.49         30.361        3815        26.27          28.869          420  
 Laptop SSD                          8K           10%         |    181.01          4.959       23168        20.22           5.049         2588
--------------------------------------------------------------+---------------------------------------------------------------------------------

To get these results I use DIskSpd tools.
With two set of parameters (64K and 8K block)

• diskspd.exe -b64k -d60 -t16 -o8 -w10 -W5 -c20G -h -L S:\DiskSpd_DeleteMe.dat > .\test_S_b64.txt
• diskspd.exe -b8k -d60 -t16 -o8 -w10 -W5 -c20G -h -L S:\DiskSpd_DeleteMe.dat > .\test_S_b8.txt
  , where S - is the disk to test

Could you run that test on your disks? with the same parameters? It take two times per 60 seconds.

Thanks!

[darklight]

Well, I have no possibility to test Azure speeds right now, however the results you shared are quite OK.
How are your SYNC/iSCSI networks configured? Bandwidth? Is it possible that network is bottlenecking your performance?

[Vladislav (Staff)]

Hello!

We have prepared a template which automatically deploys StarWind 2 Node cluster to Azure.

It could be a great option for you to try: https://github.com/starwinddeploy/azure-sw-cluster

Looking forward for your feedback.

[vmoskalenko]

Well, I have no possibility to test Azure speeds right now, however the results you shared are quite OK.

Could you test not-Azure environment? Even not-StarWind environment is ok to me.
Yes, results looks OK, but not great.
Actually, I have got the message today: "High Availability Device iqn.2008-08.com.starwindsoftware:data2, command "READ". Underlying device response time is longer than expected. Response time is 32891 ms."

How are your SYNC/iSCSI networks configured? Bandwidth? Is it possible that network is bottlenecking your performance?

I use one NIC. I have to use only one NIC. And I know it's not OK for Failover cluster and it's not OK for StarWind.
Yes, Network might be a bottleneck. I have got up to 2Gbit/s

[vmoskalenko]

We have prepared a template which automatically deploys StarWind 2 Node cluster to Azure.
It could be a great option for you to try: https://github.com/starwinddeploy/azure-sw-cluster

Where have you been month ago? when I start building 2 node cluster...
Sorry, I have not use your script.

Have your script use this https://gallery.technet.microsoft.com/s ... r-7c207d3a ?
That script really help me. And avoid me from creating Node A, then Node B, then changing Cluster Network Settings.

[Tarass (Staff)]

Sorry to hear that. Have you contacted our support team regarding your setup (and script) at initial configuration phase?

Thanks for sharing the WSFC Cluster Creation script link to the Community. Might be very useful.

[vmoskalenko]

Sorry to hear that. Have you contacted our support team regarding your setup (and script) at initial configuration phase?

Sure
I'm working with Andrii, Max and Anatoly.

Also, I PM to Vladislav with some comments about Azure PowerShell script, but he doesn't respond to me yet.

[Vladislav (Staff)]

Hi there,

Thank you for your feedback on Azure deployment template.

Sorry for the delay, I saw your message and have already informed my colleague about your recommendations.

Right now I'm working with Andrew on another request of yours