(01-11-2013, 06:06 PM)epixoip Wrote: [ -> ]Regarding your configuration: are you using multiple network interfaces on each system in an attempt to multi-home them?
No, all hosts are using one interface and one IP address so far.
do you have more than one copy of opencld running on any of your nodes?
Pretty much like a httpd, there is only one 'mother' opencld-1.2 running on each of the workers. When oclHashcat runs, there is another one spawned for the session. That one seem to live throughout the session (pid doesn't change. Also, I haven't seen the tcp source port change - there is just one long TCP session).
One thing hit me now: I've been using vclrun lately, as opposed to LD_LIBRARY_PATH=/usr/lib/vcl. It is the "same" thing, no? I am almost 100% sure I have tried both with no difference.
it's the same thing.
can you do further testing to determine if you're only seeing this with certain algorithms, certain modes, certain number of hashes, plus vs lite, etc?
It seems I was fairly lucky picking md5crypt for my experiment. I didn't have much time now but I actually could not find any other hash type that works at all, neither with lite or plus. Within 5 secs, I get "ERROR: clEnqueueReadBuffer() -38", ie. CL_INVALID_MEM_OBJECT.
I'll wait for vclHashcat before spending more time on this.
Yes, that is the error you will receive if you are not using the vclHashcat binaries.
Maybe atom can do another quick release, or give you a beta license so that you can continue testing.
VCL 1.18 is out. Anyone tried it?
Good to know, thanks. I will update the vclHashcat binaries for VCL 1.18.