This appears to be an ESX Hypervisor crash. CouchDB doesn't do anything
that any other app can do in terms of crashing your hypervisor. Probably
a VMWare bug, so check your host server logs and see if you can isolate
what is causing the crash. IMO, this has nothing whatsoever to do with
CouchDB, other than it happened to be the application putting load on
the server at the time.
Wendall
On 05/22/2012 02:25 PM, Daniel Myung wrote:
Hi folks,
I built couchdb 1.2.0 from source on our RHEL6 machine. We're looking to
transition our production environment to this new machine once it's fully
setup. However we're running into serious stability issues with the setup.
When under moderate load (like building views on one design doc that's
pretty lightweight - no reduces, just 4-5 views emitting simple property
for a key, and null values), the machine will unexpectedly reboot. The
system log files and dmesg show no sign of distress, duress, or panic, just
as if it's humming along, then the logs show the standard bootup messages
with no indication of an orderly shutdown.
The top processes shown at the time show regular CPU/memory usage of
couchjs, beam, then bam, the entire machine gets borked hard.
The total DB size is 12 gigs, and we're in the process of splitting it out
to smaller chunks. It's been a rather frustrating process on this install
as view index rebuilds would cause it to crash, and most recently, a local
filtered replication hung and caused the machine to reboot.
*System Info:*
RHEL 6.1 (stock, latest updates as of this writing, nothing else running
right now)
2.6.32-131.28.1.el6.x86_64 #1 SMP Thu Apr 19 06:41:00 EDT 2012 x86_64
x86_64 x86_64 GNU/Linux
The machine is running off of vmware ESX hypervisor, with 3CPUs and 6gb of
ram.
The only other config of note I can think of is that our couchdb databases
and views reside in directories mounted via ecryptfs.
Have others ran into issues with RHEL6 and 1.2.0?
Thanks,
Dan
