Forum Stats

  • 3,734,166 Users
  • 2,246,899 Discussions
  • 7,857,145 Comments

Discussions

NDB instance take (too) long to recover

user3915916
user3915916 Member Posts: 1
edited Feb 21, 2019 6:40AM in MySQL Community Space

Hi NDB experts,

I have a question concerning MYSQL NDB, the in memory solution.

Question is technical, not sales related.

Setup:

in most cluster configurations we setup we chose 2 NDB instances, 2 MGM's,2 arbitrators and 2 or more mysqld instances.

The issue:

When the "primary" (President) NDB goes down the remaining NDB instance takes full control.

We noticed that sometimes this happens in "seconds", at other moments this takes minutes (even upto 40 mintes).

the time used is required to recover the database.

This is evident from the NDB logs.

Obviously this takes unacceptable long down time.

Personally I believe this is due to :

LCP - GCP not "in sync", with respect to the remaining NDB instance, when the primary goes down.

I'm not able however, to find "how this works" (and with this if i'm correct)

A better understanding of this mechanism may lead to better tuning or setup with that avoiding the long down time.

The question:

Is there any document available that provides the answer ?

Many thanks in advance

Walther Spitters

Answers

Sign In or Register to comment.