top button
Flag Notify
    Connect to us
      Site Registration

Site Registration

How to improve mysql's query speed in table 'token' of keystone, when using in openstack with lots of VMs?

+1 vote
113 views

When I used mysql as the keystone's backend in openstack, iI found that the 'token' table saved 29 millions record (using myisam as engine,the size of token.MYD is 100G) and have 4 new token save per second. That result to the slow query of a token .since of inserting new token frequently, how could I set the configure to speed up the query operation.

the token's struct is id,expires,extra,valid,user_id with index {expires,valid}
and the select sql is select id,expires,extra,valid,user_id from token where valid=1 and expires >='XXXX-XX-XX XX:XX:XX' and user_id ='XXXXXXXXXXXXXXXXXXX';with often return 2 results.

Here is some db status data in a real openstack environment with 381 active VMs:

+-----------------------+-------------+ 
| Variable_name | Value | 
+-----------------------+-------------+ 
| Handler_read_first | 259573419 | 
| Handler_read_key | ********** | 
| Handler_read_next | ********** | 
| Handler_read_prev | 1235 | 
| Handler_read_rnd | 1951101 | 
| Handler_read_rnd_next | ********** | 
+-----------------------+-------------+ 

and

+-------------------------+------------+ 
| Variable_name | Value | 
+-------------------------+------------+ 
| Qcache_free_blocks | 498 | 
| Qcache_free_memory | 1192512 | 
| Qcache_hits | ********** | 
| Qcache_inserts | 352700155 | 
| Qcache_lowmem_prunes | 34145019 | 
| Qcache_not_cached | ********** | 
| Qcache_queries_in_cache | 1681 | 
| Qcache_total_blocks | 4949 | 
+-------------------------+------------+

it seems that the 'insert' operation of saving new token affects the query buffer,and result of a low-level of query-hit's rate.

Please help me?

posted Aug 2, 2014 by anonymous

Looking for an answer?  Promote on:
Facebook Share Button Twitter Share Button LinkedIn Share Button

Similar Questions
+1 vote

In VEPC, in openstack I created three VM's and when I pump data using iperf, client is sending data in application layer and the data is going through our LTE stack in receiver side it is reaching till UE ( downlink ) but iperf server is not showing the receiving data, why?
Is it a problem with mtu size? What should I change in openstack?

+2 votes

What is the difference between:

  1. using "the Command-line interface to the OpenStack APIs" through the command "openstack"
    and
  2. using the separate APIs of the different components (for example using the command-line interface to the Neutron APIs directly)?

Advantages and disadvantages?

...