Hi,
Are the reads done with autocommit=1 and outside of transactions?
All reads inside transactions will be routed to the current
master. You can do a sanity check with the command line client and
execute SELECT @@hostname, @@server_id
to see where the query is routed. You can also enable log_info in
MaxScale to see detailed logging about routing decisions.
Markus
--
You received this message because you are subscribed to the Google Groups "MaxScale" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxscale+u...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/ae813888-e591-4d44-94db-e61afeadd5c1n%40googlegroups.com.
-- Markus Mäkelä, Senior Software Engineer MariaDB Corporation
Hi,
I would recommend adding log_info=true under the [maxscale] section of your configuration and capturing the log output from a single session. It's hard to tell why the routing is behaving like it currently does without any of the log output.
You can also test if removing any of the extra parameters you defined helps with the issue:
Markus
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/90160cae-f21c-4b11-8a67-2b9fd465cffcn%40googlegroups.com.
Hi,
MaxScale seems to be working correctly. The server2 is the current master: Selected Master: server2
You can check which server is the master with: maxctrl list servers
Markus
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/00af784b-ca97-417e-b412-97910c3ca944n%40googlegroups.com.
Hi,
That might be due to your use of slave_selection_criteria=ADAPTIVE_ROUTING, it's designed to direct traffic to a single node until it starts to slow down and it's not completely deterministic as it picks a server based on the likelihood of it being the fastest to respond to it. It's the inverse of the default slave_selection_criteria=LEAST_CURRENT_OPERATIONS which tries to balance the load as equally as possible based on the number of queries being executed. Under high loads both load balancing methods end up with very similar results.
If you want to test how the queries and connections are load
balanced, I'd recommend testing with multiple client connections
in order to adequately saturate the servers with queries. This
will force the load balancing to actually do some work instead of
always picking the server with the least ongoing queries.
Markus
There are queries A, B, C, but query A goes only to server2, and query B is routed only to A. Is there a reason that queries are only routed to a specific server?
You received this message because you are subscribed to the Google Groups "MaxScale" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxscale+u...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/d4f8190a-a7c4-4a58-a757-b6608b67b092n%40googlegroups.com.
Hi,
Without knowing more about your installation and the type of traffic going through MaxScale, it's hard to say why server3 isn't being used. You might want to check out what maxctrl list servers says about it: the connections and active query counts should be distributed equally for all three servers.
Markus
--
You received this message because you are subscribed to the Google Groups "MaxScale" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxscale+u...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/28fe25d7-2990-4a42-a686-f8416f4f636cn%40googlegroups.com.
Hi,
My apologies, I was confusing the output with the old MaxAdmin command which did show it. You should see it in the show servers output which also has statistics about the server usage.
Markus
To view this discussion on the web, visit https://groups.google.com/d/msgid/maxscale/e3becf57-3fc7-4524-9f33-752534a6c736n%40googlegroups.com.