HPE Storage Users Group
https://3parug.com/

Excessive RCIP TCP retransmits
https://3parug.com/viewtopic.php?f=18&t=911
Page 1 of 1

Author:  nsnidanko [ Fri Sep 12, 2014 9:27 am ]
Post subject:  Excessive RCIP TCP retransmits

Does anyone know how 3par calculates/detects these? We replicate over very reliable P2P WAN link with 1-1.5ms latency. We checked interfaces on our network equipment and neither one shows any errors. Additional we have other applications, which pump massive amount of data and they don't have any issues.
Quote:
Event urgency: alert
Event count: 1
Event location: Site
Event time: 2014/09/11 10:34:57.00 (-0700 PDT)
Event description: 3PAR INSERV evt_tcp_excessive_retransmits

Abstract:
Excessive TCP retransmits at 9.6% on node 0. (Degraded,evt_tcp_excessive_retrans

Text:
Event id: 8254192 Node 0 Cust Alert - Yes, Svc Alert - Yes Severity:
Degraded
Event time: Thu Sep 11 10:34:57 2014
Event type: evt_tcp_excessive_retransmits Alert ID:
169 Msg ID: 1a0014
Component: Node 0
Short Dsc: Excessive TCP retransmits at 9.6% on node 0.
Event String: Excessive TCP retransmits at 9.6% on node 0
.

Author:  hdtvguy [ Fri Sep 12, 2014 11:59 am ]
Post subject:  Re: Excessive RCIP TCP retransmits

RCFC or RCIP?

We have been fighting this battle for coming up on 14 months now and have all levels of 3par support and engineering involved. The short answer after all our fighting is that 3par replication is crude and not very flexible. It can not handle congestion in the network in any way, it starts going into re-transmit way too easily. The way it has been explained to us is that the 3par watches the RC buffers and if they do not deplete or shrink in a certain amount of time the array assume a network issue and re-transmits. We have been trying to get a more definitive answer to this and actually am working for 4 engineers today as they try various tuning parameters in the 3par. 3par hates when I say this, but replication and RC were a long afterthought and a huge weakness in their technology. The majority of our issues and concerns always revolve around RC and the inefficiencies and issues with trying to keep our arrays in sync.

Author:  bajorgensen [ Sun Sep 14, 2014 5:24 am ]
Post subject:  Re: Excessive RCIP TCP retransmits

Could you contact me please?
We have an issue with remote copy that seems similar to you.
Been working on it for 12 months and while engineering has sort of acknowledged the problem,
there are very little progress on this.
I'd very much would like to share my report and discuss with you.
We have some leverage on HP and are getting to the point we have to use it.

My email is bjorn.jorgensen att basefarm.com

Author:  nsnidanko [ Mon Sep 15, 2014 7:52 am ]
Post subject:  Re: Excessive RCIP TCP retransmits

hdtvguy,

We have RCIP via 1Gbps Ethernet link. I didn't open the ticket with HP yet knowing how frustrating support can be I figured I'll ask the question here first.

How do you guys deal with these alerts? My mailbox gets emails daily and I was wondering if there is a way to disable monitoring for specific Event type.

Author:  hdtvguy [ Mon Sep 15, 2014 10:41 am ]
Post subject:  Re: Excessive RCIP TCP retransmits

We set up suppression alerts in the SP so that we do not get them any more.

Author:  hdtvguy [ Mon Sep 15, 2014 10:44 am ]
Post subject:  Re: Excessive RCIP TCP retransmits

I have zero confidence in anything they release until MU2 or MU3 are out. We consistently get burned by issue sin the GA code or their releases. We implemented 3.1.3 and had node crashes and then the false thermal issues. They were supposed to be fixed in 3.1.3 MU!, but only the node crashes seem to be fixed. We are going to stay where we are until they get they act together. Too many new features coming so I will wait until they are actually working well. Also the array will only use up to 2TB of SSD for flash cache, depending on model.

Page 1 of 1 All times are UTC - 5 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/