> So, a couple of things. First, the ARC underflow/overflow shift isn't showing > zfs_arc_overflow_shift (which is a constant) but rather the expression that > is evaluated against zfs_arc_overflow_shift: the degree to which the ARC is > overflowing (or underflowing) in terms of a power-of-two delta.
Thanks, that's what I was confused about. > Third, it's important to realize that the ARC's reaping of its own caches is > always asynchronous: kmem reaping happens out of the kmem_taskq, so the only > question is whether or not the ARC reclaim is blocking itself on this > asynchronous activity. Agreed. I phrased it poorly, but this is what I meant. > To me, our fix rectifies this underlying problem while introducing the least > amount of new complexity to the system -- an important consideration given > that it is the system's very complexity that has created this (long-standing) > issue. @ahrens since I wasn't involved in the design of the fix we adopted, can you comment on which you prefer? I'll defer to you and @bcantrill on which we should adopt for illumos. -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: https://github.com/openzfs/openzfs/pull/486#issuecomment-341516776 ------------------------------------------ openzfs-developer Archives: https://openzfs.topicbox.com/groups/developer/discussions/Tf18bbbd46b0af4a7-M17aed00ee071469ba5091cfd Powered by Topicbox: https://topicbox.com