Ask Your Question
1

Is it safe to ignore the "Service Unavailable" errors when running Swift ssbench?

asked 2014-06-12 02:45:04 -0500

kype gravatar image

updated 2014-06-13 16:58:58 -0500

smaffulli gravatar image

I have just done a fresh installation of RDO PackStack and am using SSBench to benchmark the Swift storage system. When i run this command to start the benchmarking process

ssbench-master run-scenario -f scenarios/very_small.scenario -u 4 -c 80 -o 613 --pctile 50 --workers 2

During the run many exceptions are raised:

WARNING:calculate_scenario_stats: exception from worker 1: ClientException('Object GET failed',)
INFO:Traceback (most recent call last):
  File "/usr/lib/python2.6/site-packages/ssbench/worker.py", line 206, in handle_job
    handler(job_data)
  File "/usr/lib/python2.6/site-packages/ssbench/worker.py", line 444, in handle_get_object
    resp_chunk_size=object_info.get('block_size', DEFAULT_BLOCK_SIZE))
  File "/usr/lib/python2.6/site-packages/ssbench/worker.py", line 357, in ignoring_http_responses
    raise error
ClientException: Object GET failed: http://192.168.36.229:8080/v1/AUTH_a25f1a2f28a74907a06532147633c020/ssbench_000001/small_000001 503 Service Unavailable  [first 60 chars of response] <html><h1>Service Unavailable</h1><p>The server is currently

However SSBench is still able to get me the results of the benchmark

Small test scenario  (generated with ssbench version 0.2.23)
Worker count:   2   Concurrency:   4  Ran 2014-06-01 19:13:27 UTC to 2014-06-01 19:13:30 UTC (2s)

% Ops    C   R   U   D       Size Range       Size Name
 91%   % 10  75  15   0        4 kB -   8 kB  tiny
  9%   % 10  75  15   0       20 kB -  40 kB  small
---------------------------------------------------------------------
         10  75  15   0      CRUD weighted average

TOTAL
       Count:   101 (  512 error;  4718 retries: 4671.29%)  Average requests per second:   7.1
                            min       max      avg      std_dev  50%-ile                   Worst latency TX ID
       First-byte latency:  0.016 -   0.273    0.053  (  0.031)    0.047  (all obj sizes)  txe207b1a88a37420aa12b8-00538b7b5a
       Last-byte  latency:  0.016 -   0.274    0.053  (  0.031)    0.047  (all obj sizes)  txe207b1a88a37420aa12b8-00538b7b5a
       First-byte latency:  0.016 -   0.161    0.051  (  0.022)    0.047  (    tiny objs)  txd876e69941574f26acbac-00538b7b5a
       Last-byte  latency:  0.016 -   0.161    0.051  (  0.022)    0.047  (    tiny objs)  txd876e69941574f26acbac-00538b7b5a
       First-byte latency:  0.027 -   0.273    0.075  (  0.071)    0.059  (   small objs)  txe207b1a88a37420aa12b8-00538b7b5a
       Last-byte  latency:  0.028 -   0.274    0.076  (  0.071)    0.059  (   small objs)  txe207b1a88a37420aa12b8-00538b7b5a

READ
       Count:   101 (  350 error;  3509 retries: 3474.26%)  Average requests per second:   7.1
                            min       max      avg      std_dev  50%-ile                   Worst latency TX ID
       First-byte latency:  0.016 -   0.273    0.053  (  0.031)    0.047  (all obj sizes)  txe207b1a88a37420aa12b8-00538b7b5a
       Last-byte  latency:  0.016 -   0.274    0.053  (  0.031)    0.047  (all obj sizes)  txe207b1a88a37420aa12b8-00538b7b5a
       First-byte latency:  0.016 -   0.161    0.051  (  0.022)    0.047  (    tiny objs)  txd876e69941574f26acbac-00538b7b5a
       Last-byte  latency:  0.016 -   0.161    0.051  (  0.022)    0.047  (    tiny objs)  txd876e69941574f26acbac-00538b7b5a
       First-byte latency:  0.027 -   0.273    0.075  (  0.071)    0.059  (   small objs)  txe207b1a88a37420aa12b8-00538b7b5a
       Last-byte  latency:  0.028 -   0.274    0.076  (  0.071)    0.059  (   small objs)  txe207b1a88a37420aa12b8-00538b7b5a

So is it safe to ignore the Python exceptions raised above? Thanks in advance

My Setup:

PackStack RDO All-IN-ONE configuration (Icehouse release) running on CentOS 6.5

SSBench installed on same system

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2014-10-07 15:27:04 -0500

It's not unheard of for a benchmark of this kind to result in 503 Service Unavailable responses, simply because the service is being put under stress and isn't able to keep up. It's not ideal, obviously, but I don't think it's something to be concerned about. So, yes, I believe it's safe to ignore them.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

[hide preview]

Get to know Ask OpenStack

Resources for moderators

Question Tools

Follow
1 follower

Stats

Asked: 2014-06-12 02:45:04 -0500

Seen: 541 times

Last updated: Oct 07 '14