Jun 26, 2008 7:42 AM
Zenoss Performance and Latency tuning
-
Like (0)
ProxyPass / http://localhost:8080/ retry=5
ProxyPassReverse / http://localhost:8080/
CacheEnable mem /
vm.vfs_cache_pressure = 5
vm.swappiness = 5
vm.overcommit_memory = 2
vm.dirty_background_ratio = 5
vm.dirty_ratio = 80
vm.dirty_writeback_centisecs = 3000
vm.dirty_expire_centisecs = 9000
net.core.rmem_max=16777216
net.core.wmem_max=16777216
net.ipv4.tcp_rmem=4096 87380 16777216
net.ipv4.tcp_wmem=4096 65536 16777216
net.ipv4.tcp_no_metrics_save=1
net.ipv4.tcp_congestion_control=bic
But mod_proxy alone doesn't add any performance gains and the same
functionality could be gained using Apache rewrite rules. What can
improve latency is using mod_proxy in combination with mod_cache and
memory web caching:
Code:
CacheEnable mem /
ProxyPreserveHost On
I guess people want to be polite in the christmas spirit ;-)
But I had the same experience while reading the text "uhh... wtf?" :-)
Regarding performance I have been playing with the mysql settings in zenoss...
Edited /usr/local/zenoss/mysql/my.cnf and added (below [mysqld]):
bind-address = 127.0.0.1
key_buffer_size = 32M
sort_buffer_size = 1M
read_buffer_size = 1M
read_rnd_buffer_size = 1M
join_buffer_size = 1M
myisam_sort_buffer_size = 1M
max_allowed_packet = 16M
net_buffer_length = 8K
table_cache = 2048
thread_stack = 192K
thread_cache_size = 8
thread_concurrency = 8
query_cache_size = 32M
query_cache_limit = 1M
query_cache_type = 1
ft_min_word_len = 2
ft_stopword_file =
skip-external-locking
skip-name-resolve
default-storage_engine = innodb
transaction-isolation = READ-COMMITTED
innodb_locks_unsafe_for_binlog = 1
innodb_additional_mem_pool_size = 16M
innodb_buffer_pool_size = 256M
innodb_doublewrite = 0
innodb_flush_log_at_trx_commit = 2
innodb_flush_method = O_DIRECT
innodb_log_buffer_size = 8M
innodb_log_file_size = 128M
innodb_log_files_in_group = 2
innodb_rollback_on_timeout = 1
innodb_thread_concurrency = 8
and added (below [mysqld_safe]):
nice = -5
open-files-limit = 8192
Perhaps someone with a bit of sparetime who wish to confirm the above settings?
Afterwards restart the mysql service:
/etc/init.d/zenoss-stack restart mysql
If error occurs during startup you might need to delete the innodb logfiles (ib_logfile0 and ib_logfile1) located in /usr/local/zenoss/mysql/data (due to changed innodb_log_file_size) and then:
/etc/init.d/zenoss-stack start mysql
In order to access mysql from cli you can do this (to login as root, otherwise change the -u):
/usr/local/zenoss/mysql/bin/mysql -S /usr/local/zenoss/mysql/tmp/mysql.sock -u root -p
The default password is "zenoss" unless you have changed it to something else during install.
Regarding the scheduler I prefer "noop" for servers and this can be changed during runtime:
echo "noop" > /sys/block/sda/queue/scheduler
current setting (and options) can be disabled by:
cat /sys/block/sda/queue/scheduler
Our Zenoss installation was unusable. We've added an additional collector at the data center, but that didn't help much. Today I've increased the ZODB cache, which increased the performance significantly.
First, you will need to check the number of entries that you have in your catalog, by simply going to the link below :
http://yourdomain:8080/zport/global_catalog/manage_catalogView?submitted=true
Round it up to the nearest 100,000.
Edit your zope.conf
locate the section below and increase the cache to the rounded number
then Restart your zenoss-stack, and you will see a huge performance increase.
Follow Us On Twitter »
|
Latest from the Zenoss Blog » | Community | Products | Services Resources | Customers Partners | About Us | ||
Copyright © 2005-2011 Zenoss, Inc.
|
||||||||