Print

Print


Hi Rachid,

> What is the impact of the loss of a node on qserv ?
If the node is fully down, the impact is loss of the data on that node. 
This means that queries that depend on the lost data will fail. Some 
queries can complete if they are completely independent of the lost 
data. In the current data distribution scheme, each node is responsible 
for data portions spread over the entire data set.

> Can the service be recovered by excluding the node ?
An administrator can reconfigure the master node to exclude the lost 
portions of the data set, and restart the master (losing in-flight 
queries, causing inconsistent state on workers). This removes the lost 
portions from the data set.

Currently, there is no way for a qserv master to contact its workers to 
reset them.

If there is replication, service should continue normally, except that 
in-flight queries that were assigned to the node will fail. New queries 
can complete, as long as there is still a full copy of the data set.

Hope this helps,
-Daniel

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1