I have discovered an issue when trying to store a lot of keys using python-memcached.
Here is what I’m doing:
import memcache
# This dict has a 2270 entries that is generated on production server
v = eval(open("/home/dragoon/output").read())
a = memcache.Client(['unix:/tmp/memcached.sock'], debug=1)
Next I’m trying to set all these values:
In [94]: len(a.set_multi(v))
MemCached: MemCache: unix:/tmp/memcached.sock: timed out. Marking dead.
Out[94]: 2270
The log output of the memcached server:
...
27: going from conn_new_cmd to conn_parse_cmd
<27 set d750bde63a98579f9c2987907aaaf5f8 1 0 18
27: going from conn_parse_cmd to conn_nread
> FOUND KEY d750bde63a98579f9c2987907aaaf5f8
>27 STORED
27: going from conn_nread to conn_write
Failed to write, and not due to blocking: Broken pipe
27: going from conn_write to conn_closing
<27 connection closed.
After some time:
In [96]: len(a.set_multi({'test':1}))
Out[96]: 0
In [97]: a.get('test')
Out[97]: 1
Default timeout is 3 seconds, but increasing it even to 100 seconds doesn’t help, memcached is just getting stucked.
So, my question is, what is the problem with memcache?
I saw people talking about storing over 1000 000 entries in it, but for me it can’t even store 2 000?
Ok. It seems that there is some bug in memcached 1.4.2 that is the last version in Ubuntu 10.04 LTS.
Memcached 1.4.5 works fine.