maxmerges too small: is this an error ?

Luc_Lecocq

Hi,

On our test cluster running 7.1.0, we continuously see in the vertica logfile messages saying:
2015-02-23 14:46:29.890 Init Session:0x7fafdc0135c0-a000000098de01 [EE] <INFO> SortManager found maxMerges 7 too small(0 MB Assigned).
2015-02-23 14:46:29.890 Init Session:0x7fafdc0135c0-a000000098de01 [EE] <INFO> After disabling optimization, maxMerges becomes 7.

Can someone explain what this means and what are the consequences ?
I don't know if this is relevant, but at the same time, we observe that many projections remain for days in state "being reorganized", without any progress on the percentage, between 0 and 75%  ?

Thanks for helping,

[Deleted User]

Hi Luc,
This is just a heads up that we are migrating this Community to a new platform in a couple of days. We’ll make every effort to answer your question before that…but if it remains open, please do visit our new forum and pose your question once again. You’ll find the new forum here: https://community.dev.hp.com/t5/Big-Data-and-Analytics/ct-p/bigdata_analytics. Please visit us there on Thursday!

Rob_Vugts

I'm seeing exactly the same messages on version 7.1.1-3:

 

 [EE] <INFO> SortManager found maxMerges 7 too small(0 MB Assigned).
 [EE] <INFO> After disabling optimization, maxMerges becomes 7.

yardenbar

Hi All,

Any update/solution on this subject?

I'm haing the same issue on my cluster

 

Thank you,

Yarden

Rahul_Choudhary

Hi

 

Just to let you know that maxMerges message typically indicates that  how many sorted containers we try to merge at once. Merging more containers takes more system resources and can potentially slow down each merge phase, but it can also reduce the total number of merge phases.

This parameter cannot be tuned directly; if you want to influence it, the correct way to do so is by assigning more or less memory to the query via resource pools.

 

What you're seeing here is debugging output, used to verify that the computation is correct.

 

Thanks

Rahul