Damir Krstic
2017-07-05 13:29:18 UTC
We have recently upgraded all our compute nodes to RHELS 7.3. We have left
the management node on RedHat 6. After the nodes were upgraded and booted,
we are experiencing a slow down on our private (172.20) network. For
example, psh compute date command finishes successfully and quickly, but
when we try to copy password file across all nodes, it hangs almost always
on different nodes. I thought I had this issue isolated to a specific rack,
but now it's obvious that it happens on all nodes. As a workaround, I am
copying files across the cluster using -ib0 interface.
I was wondering if RH7 is doing something funky with routing, or
networking. I have looked at the tcpdump and can't really see anything
strange except to say that it takes a long time for the packet to come back
to the management node.
I am thinking it's either routing or dns (reverse lookup) issue but I can't
be sure. I am hoping somebody on this listserv had a similar issue and was
able to resolve it.
Thanks in advance.
Damir
the management node on RedHat 6. After the nodes were upgraded and booted,
we are experiencing a slow down on our private (172.20) network. For
example, psh compute date command finishes successfully and quickly, but
when we try to copy password file across all nodes, it hangs almost always
on different nodes. I thought I had this issue isolated to a specific rack,
but now it's obvious that it happens on all nodes. As a workaround, I am
copying files across the cluster using -ib0 interface.
I was wondering if RH7 is doing something funky with routing, or
networking. I have looked at the tcpdump and can't really see anything
strange except to say that it takes a long time for the packet to come back
to the management node.
I am thinking it's either routing or dns (reverse lookup) issue but I can't
be sure. I am hoping somebody on this listserv had a similar issue and was
able to resolve it.
Thanks in advance.
Damir