Date: Thu, 3 Nov 2016 14:31:37 -0400 From: Prakhar Goel <newt0311@gmail.com> To: freebsd-cloud@freebsd.org Subject: FreeBSD images swap space on GCE Message-ID: <CABa1M22%2B%2BOUBh6vuoOrBo-q-RV-osz7EcVeWt47fB0ck2id5UQ@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
Hello, Something I came across recently: I'm using freebsd-11-0-release-amd64 from the freebsd-org-cloud-dev for some of my VMs. Under high memory pressure they seem to just lock up and become unresponsive. I've waited for up to ~15 minutes for them to revive with no luck. After a little debugging, I think I've narrowed the problem down to the swap space. By default, the VM seems to have a 1GB swap space setup on the root disk. GCE scales I/O linearly with the size of the disk and the size of the root disk is only 22GB. I think this leaves the root disk with too little I/O capacity to act as swap space (it's still enough for normal fs use apparently). When I was monitoring this in top, my memory-intensive processes would continue to work until the system ran out of RAM and then they'd just get stuck in "vmwait." To resolve the problem, I enabled extra space in an attached volume (standard persistent disk, 200GB size). As long as my new swap space was enabled *and the old swap space was disabled*, my applications were fine. Not sure what can be done to resolve this. The default swap setup just seems to cause problems so maybe remove it altogether and let users set up their own swap space on faster devices as needed? Thanks. PS. Is this the right place to send emails like this? Not sure myself. -- ________________________ Warm Regards Prakhar Goel
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CABa1M22%2B%2BOUBh6vuoOrBo-q-RV-osz7EcVeWt47fB0ck2id5UQ>