changing node X startup state to critical

Pietro_La_Torre

Hi,
I have a Vertica 6.1.2 single-node cluster which very often writes the following message on Vertica MC:

changing node X startup state to critical

By going in detail through Vertica doc, I found that a node is in critical state when its failure would cause a DB shutdown since there is no K-safety. Of course, since I have only 1 node, I already know that I don't have K-safety.

I would like to troubleshoot this issues.

Are there any other reasons that could cause this message ? Does it mean that I have bad issues or I could just ignore them ?
Also..
why is this message shown periodically?
Is it possible to disable this message ? 

Thanks,
Pietro

[Deleted User]

Hi Pietro,

Vertica's durability and crash-recovery depend on K-safety.  If a node has committed data in the WOS and that node crashes, the only way to get the data back is to retrieve it from another node.  This is a fundamental limitation of running Vertica without K-safety (and therefore of Vertica as a single-node database); also a limitation that we find that people often forget, and re-discover only after a node has crashed.

The MC is targeted at displaying issues that likely matter to production services.  We allow single-node Vertica instances because they are great for development purposes.  But I would not personally recommend running a production service on a single-node database.  (Use three small machines instead.  We have a whitepaper somewhere about running Vertica on a cluster of many HP Moonshot blades, for example; tiny cheap Intel Atom-based machines.)

I don't know whether there's a way to disable the warning.  Others may be able to comment there.

Adam

Pietro_La_Torre

Hi Adam,
thank you for your reply.
Just to be clear, do you confirm that this messages could be displayed just for warning and they could be not directly related to any issues?
My only node haven't crashed yet, but I often see this message

Pietro_La_Torre

furthermore, I have other development environment with single-node clusters but MC does not display this message in those cases. I can't figure out why.

[Deleted User]

Hi Pietro,

Many Vertica users would consider "the current configuration will cause you to lose data if the system crashes" as an issue.  (And a serious issue, too.)

The error doesn't indicate that your system is more likely to crash, if that's your concern.  Just that the cost is greater if it does.

Adam