We have been testing out iSCSI on a ESX4 host and have noticed a problem.
At this point we do NOT have any host running off of iSCSI. There are two host that are on the iSCSI datastore but they are powered OFF.
If we power off our iSCSI target (an openfiler server) all of the VMs on the server appear to hang for at least few seconds up to a minute at a time at least twice an hour. We aren't sure if they are actually hanging or if the networking just stops working. During that time period we can't run esxtop because the service console also stops responding. We have not been logged into the console directly when it happens to prove if the entire host is hanging or just the network stack. As soon as we turn our iSCSI server back on everything goes back to normal. The problem started when we started testing iSCSI and we can't reproduce the issue when our iSCSI server is live.
During the outage time we also get gaps in our performance data. On our CPU percent graphs we see one core spike to 100% right before the gap but on the VM usage graph we don't see the spike (see attached).
Anyone have any ideas what could be causing this or how to fix the problem? We plan on turning off iSCSI but that requires a reboot of the EXS server.
