Fully managed cloud. 30-day FREE full featured trial. Start Now
cancel
Showing results for 
Search instead for 
Did you mean: 

IDEA spotlight 01/2022

mwellbro
Xpert

Hello everyone,

and welcome back to the "IDEA spotlight series", where I read ideas so you don´t have to 😉

Seriously though, I hope you all had a great start into 2022 and we´re all looking forward to the nice new things we
may get our database system to do for us.

Speaking of which, here´s this months spotlighted IDEA, this time taken from the IDEAs portion of the community by checking
which IDEAs reside at the end of the "Top Ideas"-list but might deserve some attention nonetheless:

https://community.exasol.com/t5/ideas/realtime-switchover-to-reserve-nodes-migrated/idi-p/4460


Having the database restart to switch over to a reserve node is a short interruption, but it´s still an interruption and might cause a lot of grievances down the line - I´m somewhat sure that a handover from the failed node to the reserve node where we´d keep the exact mem-contents 1:1 isn´t really feasible ( if at all possible ) but I could see a systematic approach that might 

1) get all active SQL-texts from all active sessions to a surviving node
2) signal a transient failure to the clients ( this might get tricky since as far as I remember the client has one "main connected node" which in case of the failed node wouldn´t be in place to signal the transient failure ) and ask the driver to restart the current statement in a few seconds
3) once the reserve node is up the surviving node distributes the SQL-texts among the cluster nodes and restarts them
4) the clients will receive the result sets once the restarted SQLs are done

This probably won´t work if the statement already began to fetch data / push data to the client ( risk of sending the same data
multiple times ), and most likely for a number of other reasons ( I think the way Exa is managing it´s disks might rain on this particular parade, among other things ) - but still, this might be an interesting line of thought to explore.

How about all of you ? Are Hardware issues , Zero-downtime requirements or similar things a topic of interest for you ? If
so, maybe a vote for this IDEA might be just the thing 🙂

Cheers,
Malte 

0 REPLIES 0