I forgot to say that the limit is not just in the number of job slots. We are using a NFS server for all OSG related things, including the scratch space each Panda job will read and write. This scratch space should be batch nodes' local disk but since most of our batch nodes don't have enough scratch space, we are using the OSG NFS server for that. This will not scale. I got an answer a while ago from Xin Zhao of BNL that each job need >5GB scratch space. I am not sure where this number came from, and it is an open ended. But perhaps we should take this into consideration when we allocate batch slots and buy more machines. BTW, I redirect some jobs to griddevs. There is 20 running jobs and 23 pending now (LSF jobs). The gatekeeper is quiet so far. Wei Yang | [log in to unmask] | 650-926-3338(O) Neal V. Adams wrote: >> Should raise the limit to something like 20 or 30. >> > > Wei just dropped by and asked about this change. I've made the change > to a 30 job limit per user in the LSF configuration file and it will > go into effect during this evening's scheduled LSF reconfiguration at > approx. 19:35 PDT. > > --Neal