boost - SQL Assoc Failed

[brylant]

Hi,

I'm running boost server (latest cacti+spine+boost with all official patches) and I'm seeing this message from time to time (once or twice a day):

Code: Select all

02/04/2010 08:37:49 AM - CMDPHP: Poller[0] ERROR: SQL Assoc Failed!, Error:'2006', SQL:"SHOW TABLES LIKE '%boost%'"

Any idea what's wrong?
Here's my.cnf (I've got plenty of RAM):

Code: Select all

key_buffer              = 512M
max_allowed_packet      = 16M
thread_stack            = 128K
thread_cache_size       = 16
max_connections         = 2000
table_cache             = 512
thread_concurrency      = 4
read_buffer_size        = 2M
read_rnd_buffer_size    = 32M
sort_buffer_size        = 128M
max_heap_table_size     = 3G
tmp_table_size          = 1G
#
# * Query Cache Configuration
#
query_cache_limit       = 1M
query_cache_size        = 128M

[Deviloper]

Witness told use:

##############################################
From the MySQL Manual

Quote:
Error: 2006 (CR_SERVER_GONE_ERROR)

Message: MySQL server has gone away


MySQL is crashing. You likely are not able to create a large enough memory table due to a MySQL setting. Try the following:

Code:
mysql cacti
alter table poller_output_boost engine=myisam;


Make sure you are running MySQL 5.x

TheWitness
###############################################

[brylant]

well... i'm not sure... If that was the case I'd have MySQL crash logs - which I don't... Looks like mysql is working fine.
And I'd really like to stick to engine=memory (and not myisam)...

Any other ideas...?

[TheWitness]

Well, I warn users not to turn this up unless they know what they are doing and know the implications. One of the big ones is system memory. You need to have enough. If you have enough, you need to know how to configure MySQL to leverage it.

TheWitness

[brylant]

So (assuming I don't really know what I'm doing..) - what's wrong with my my.cnf? I've got 16GB of memory and - looking at memory stats, boost stats and mysql stats - I never reach 100% utilization...
Am I stuck with these random SQL Assoc Failures...?

[TheWitness]

So, that's odd, I wrote a response, and then re-read yours. You already have a 3G max heap table size. So, your issue could be mysql bug. They made some changes in 5.0.71++ that make the server go away from time to time. (it's a bug).

So, check for a mysql update, and then increase your thread_concurrency to say 32, and also increase your thread_cache_size to 64. That "might" help. For the user who reported this issue (actually Rony). I made both a patch to spine to allow auto-reconnect, and they increased these two variables.

TheWitness