1 DL585 64GB - 13GB used 8GB cached
Discovered 213
Network 478
Server 745
Storage 20
Most of the discovered devices are *nix that zenoss isn't monitoring yet.
Server is ~60 ESX and the rest Windows.
Windows is using one of the WMI templates and 420 of the Windows servers also
use the MySQL Zenpack.
Everything crawled along and most just didn't work once I started adding so
many devices. Now it is running a lot better.
I configured 3 monitors all on the same machine.
I had to hack the code a little, but now zencommand, zenpefsnmp, zenping,
zenstatus, zenmodler, zenprocess, zenwin, and zenteventlog all have 3 daemons
running on the 1 DL585. Without 3 processes each half of my cores were idle.
MySQL:
innodb_buffer_pool_size = 4096M
innodb_additional_mem_pool_size = 50M
innodb_flush_log_at_trx_commit = 2
table_cache = 1024
thread_cache = 16
query_cache_size = 64M
max_connections = 500 (too high. zeross only uses a few connections)
innodb_thread_concurrency = 16
zope.conf:
zserver-threads 150
default was 4 - until I changed it the web interface had to wait for an open
thread. Any wonder why you have to sit and wait while your CPU was idle?
<zodb_db main>
mount-point /
# ZODB cache, in number of objects
cache-size 150000
pool-size 50
<zeoclient>
server localhost:8100
storage 1
name zeostorage
var $INSTANCE/var
# ZEO client cache, in bytes
cache-size 250MB
# Uncomment to have a persistent disk cache
client zeo1
</zeoclient>
</zodb_db>
I was having issues waiting for zeodb also so I increased the pool-size. I
also increased the cache to hold the DB in RAM.
zenhub.conf:
#PARAMETER VALUE
cachesize 8000
pcachesize 2250
High enough cache to hold it all in RAM.
zenhub is using 1.2GB
zope is using 3.1GB
zenhub doesn't have nearly as many issues for me now. I still have timeouts
here and there and new devices sometimes don't load.
Web UI is very fast, except sometimes when I move devices into different
classes it can slow down.
polling times:
snmp command
monitor1 60s 15s
24k 925 devices
monitor2 27s 48s
11k 3k devices
monitor3 50s 22s
21k 1.3k
About 61,000 datapoints.
56,000 snmp
5,000 command
I have probably changed a few other parameters for performance, but those are
the ones off the top of my head.
[/img]
-------------------- m2f --------------------
Read this topic online here:
http://forums.zenoss.com/viewtopic.php?p=25240#25240
-------------------- m2f --------------------
_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users
