cheller [7:42 PM] So, I did some digging about the LSST compute cluster. One, it appears there are too many users thrashing it at one time and the storage can't keep up seeing that it is just one node, too many requests over too small a pipe. Two, there's a local SSD for home, but only some people are using it... It would be far better to get people to use local scratch space for compiles and heavy duty stuff vs GPFS. The people on the local storage are, of course not be effected by these data storms the users are creating. I'd suggest adding more people to home from that SSD and potentially adding more local fast scratch space if this is going to continue to be an issue. Thanks.