Home / server / Questions / 776863
In Process
Karem
Asked: 2016-05-16 00:11:14 +0800 CST

ubuntu + hhvm results in my tmpfs is full - 100% used

  • 772

Ubuntu 14.04, HHVM 3.10.1

Running df -h I can see the following:

Filesystem      Size  Used Avail Use% Mounted on
udev            7.9G  8.0K  7.9G   1% /dev
tmpfs           1.6G  1.6G  8.0K 100% /run
/dev/xvda1      9.8G  2.1G  7.2G  23% /
none            4.0K     0  4.0K   0% /sys/fs/cgroup
none            5.0M     0  5.0M   0% /run/lock
none            7.9G     0  7.9G   0% /run/shm
none            100M     0  100M   0% /run/user

What's taking up the full space in tmpfs is actually the file /run/hhvm/hhvm.hhbc which is 1.6 GB currently. Is this healthy? It does not look like it is.

I have found the following, related to the full tmpfs: https://github.com/hhvm/packaging/issues/132

"On my Ubuntu (14.04) /var/run is a tmpfs mount that's only 400MB in size. The hhvm.hhbc file grew to a little over 350MB on one system. I think it may make more sense to default to putting this file under /var/cache/hhvm instead."

My central path is just as explained, defined to /var/run/hhvm/hhvm.hhbc - but i do not have a /var/cache/ in my filesystem, so i wonder if i really should change the central path to /var/cache/hhvm instead? Also, wouldn't this be bad / not be temporary since tmpfs is mounted to only /run?

I am debugging since my HHVM is crashing periodically, which requires that i run a reboot of hhvm service/remove the hhvm.hhbc before everything works again.

ubuntu hhvm

1 Answers

  1. The hhvm.hhbc file is the hiphop compiled bytecode. From the looks of it, it's getting too big for your tmpfs. You can either change the path to somewhere on /dev/xvda1, or increase the size of your tmpfs partition (if your memory allows for it)

    Using the tmpfs path is fast, since the compiled code repo will be read from memory, setting it to an actual on disk file system will slow down the reads.

    • 0