Skip to main content

One Server crashing often / different hardware setup

  • November 16, 2015
  • 6 replies
  • 21 views

Pelle
Forum|alt.badge.img
  • Participating Frequently

Hi,

 

we run a cluster with three vertica nodes and the following hardware:

 

1. 2 x Intel Xeon 4-Core 2,40Ghz 96GB ECC DDR3 RAM

2. 2 x Intel Xeon 4-Core 2,40Ghz 96GB ECC DDR3 RAM

3. 2 x Intel Xeon 6-Core 2,40Ghz 96GB ECC DDR3 RAM

 

Is it possible that the 3rd box crashes a lot because of the different CPU?

 

I can't find anything helpful in vertica.log. The last line before the last crash was:

 

2015-11-14 07:50:06.759 DistCall Dispatch:0x7f0b5c248c40 [Txn] <INFO> Commit Complete: Txn: a00000014e4fb0 at epoch 0xc7fb76

6 replies

Chris_Daly_HPE
Forum|alt.badge.img
  • Participating Frequently
  • November 18, 2015

Pelle,

   It is certianly possible that the different configuration in Node 3 is causing issues.  When you say the box crashes alot, are we talking about vertica on the box crashing, or the entire box?

 

-Chris


Pelle
Forum|alt.badge.img
  • Author
  • Participating Frequently
  • November 18, 2015

Hey Chris,

 

thanks for your reply and sorry for being so unspecific. It's the vertica instance crashing, not the entire server.

 

Cheers


Chris_Daly_HPE
Forum|alt.badge.img
  • Participating Frequently
  • November 18, 2015

Pelle,

   Ok, in that case, what I would try first is limiting the number of cores in node 3 to match the number of cores in nodes 1 and 2.

 

Here is an article I found online that walks you through that - http://www.absolutelytech.com/2011/08/01/how-to-disable-cpu-cores-in-linux/ 

 

Typically we don't do a lot of testing on non-homogeneous configurations, so this will be interesting! :)

 

-Chris


Pelle
Forum|alt.badge.img
  • Author
  • Participating Frequently
  • November 19, 2015

Thanks, I'll try that right now :-)


Chris_Daly_HPE
Forum|alt.badge.img
  • Participating Frequently
  • November 24, 2015

I'm hoping that no news is good news.  How did restricting the number of CPUs in node 3 work out?

 

-Chris


Pelle
Forum|alt.badge.img
  • Author
  • Participating Frequently
  • December 4, 2015

Hi Chris,

 

sorry for getting back so late... I found out that all three nodes had different CPUs so we deceided to replace the hardware with identical stuff. It seems to be stable since.

 

Thanks for your help :-)