MTU size
Hi! On all of my x86_64 Linux boxes the default MTU is 1500. But with SuSE (both 11 and 12) on s390 they seems to be 1492. How come? -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
More important is too large an MSS which can cause fragmentation. You can test your network route from z/OS with this ping: ping 192.168.20.54(PMTU yes LENGTH 1472 Verbose 'PMTU yes' says 'do not fragment'. On my host 1472 is max value that does not fragment. 1473 will need fragmentation. Going over a VPN will have larger TCP/IP headers giving a smaller MSS. For dynamic detection to work, your routers must pass the necessary ICMP packets. By default, some do not. Good article here: http://www.networkworld.com/article/2224654/cisco-subnet/mtu-size-issues.html -- Donald J. dona...@4email.net On Fri, Jan 29, 2016, at 07:35 AM, Cohen, Sam wrote: > Chris, > > If you look in the hardware documentation for the Open Systems Adapter (OSA), > it has a "normal" MTU of 1492, so your Linux is matching the hardware > settings. The OSA does support jumbo frames (8992) if your network is > configured to use them. > > Thanks, > > > Sam Cohen > Levi, Ray & Shoup, Inc. > > > > -Original Message- > From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of > Christer Solskogen > Sent: Friday, January 29, 2016 1:56 AM > To: LINUX-390@VM.MARIST.EDU > Subject: MTU size > > Hi! > > On all of my x86_64 Linux boxes the default MTU is 1500. But with SuSE (both > 11 and 12) on s390 they seems to be 1492. How come? > > -- > For LINUX-390 subscribe / signoff / archive access instructions, send email > to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 > -- > For more information on Linux on System z, visit http://wiki.linuxvm.org/ -- http://www.fastmail.com - Or how I learned to stop worrying and love email again -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
On Friday, 01/29/2016 at 04:28 GMT, "Donald J."wrote: > More important is too large an MSS which can cause fragmentation. > You can test your network route from z/OS with this ping: > ping 192.168.20.54(PMTU yes LENGTH 1472 Verbose > > 'PMTU yes' says 'do not fragment'. On my host 1472 is max value that > does not fragment. 1473 will need fragmentation. Going over a > VPN will have larger TCP/IP headers giving a smaller MSS. > For dynamic detection to work, your routers must pass the necessary > ICMP packets. By default, some do not. > > Good article here: > http://www.networkworld.com/article/2224654/cisco-subnet/mtu-size-issues.html Even with all of that in place, you still need to get your local MTU value correct. It represents a ceiling on the MTU size PMTU discovery will try, and it sets the limit on the TCP MSS. Everyone connected to the same LAN segment needs to have the same MTU on that LAN segment. This is still very much a stone-knives-and-bearskins issue with TCP/IP. IMO, it's something IEEE should address since the mothership for the MTU is the local ethernet switch. That's where the MTU is configured ("Are we using jumbo frames or not?"). There's a smart device at the other end of the wire. There are a variety of things negotiated at a low level (duplex, speed, VEPA) and there's no practical reason "maximum frame size" can't be discovered. MTU for a particular protocol is a function of the MFS. (We're familiar with MFS from HiperSockets.) Alan Altmark Senior Managing z/VM and Linux Consultant Lab Services System z Delivery Practice IBM Systems & Technology Group ibm.com/systems/services/labservices office: 607.429.3323 mobile; 607.321.7556 alan_altm...@us.ibm.com IBM Endicott -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
Chris, If you look in the hardware documentation for the Open Systems Adapter (OSA), it has a "normal" MTU of 1492, so your Linux is matching the hardware settings. The OSA does support jumbo frames (8992) if your network is configured to use them. Thanks, Sam Cohen Levi, Ray & Shoup, Inc. -Original Message- From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of Christer Solskogen Sent: Friday, January 29, 2016 1:56 AM To: LINUX-390@VM.MARIST.EDU Subject: MTU size Hi! On all of my x86_64 Linux boxes the default MTU is 1500. But with SuSE (both 11 and 12) on s390 they seems to be 1492. How come? -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: brtfs -u issue
On 01/28/2016 10:26 AM, PHILIP TULLY wrote: > According to the documentation to change the uuid of a btrfs > filesystem in SLES 12 s390x, we should be using the btrfstune with > the –u option, but this doesn't seem to work. The result is we are > unable to mount a duplicate copy of a disk with BTRFS in two different > mounts on the same system. Need more detail. Can you share how you're trying to mount it? actual command sequence? Forgive me ... couldn't help but think you *might* be trying to mount-by-UUID, which would not work if the two vols have the same UUID. (i.e. stamping a new UUID on the dup vol failed) I presume you would have thought of this, but ... gotta ask. -- R; <>< -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
On Friday, 01/29/2016 at 09:58 GMT, Christer Solskogenwrote: > On all of my x86_64 Linux boxes the default MTU is 1500. But with SuSE > (both 11 and 12) on s390 they seems to be 1492. How come? Historically, it's because there are two different kinds of non-jumbo ethernet frames. The traditional "DIX" frame has room for a 1500-byte packet. The other, defined by IEEE 802.3 has a slightly longer frame header that enables multiple multi-protocol hosts to share a single network adapter, leaving room for only 1492 bytes. (See RFC 1042.) Bottom line, if you're on an IEEE 802.3 network with SNAP headers, you have to use MTU 1492. And in the Before Times, the MTU also dictated the maximum packet size a host could *receive*. (Hmmm...so other guy's MTU has to match your MRU.) But that's usually not an issue any more. Not, at least, in the mainframes -- dunno about x86 adapters. And we have dynamic path MTU discovery to help us. All of that said, as of the z13, OSAs no longer support IEEE 802.3 networks, so 1500 is the correct default from here on out. OSAs are capable of telling the host the MTU size, so Linux *can* get that value from the OSA adapter. It's just that this isn't something other platforms can do and so the configuration process for the IP layer isn't really amenable to self-discovery. All of this fancy local MTU discovery was based on ARP. When you ARP for a IP address, you include "hardware type" in the ARP packet. There is a value for 802.2 (which covers 802.3) and a value for Ethernet (DIX). First you ARP with hw=802.2. If you get a response, you use MTU 1492. If you don't you try again with hw=Ethernet. If you get a response, 1500 it is (unless you have configured for Jumbo frames). This same issue drives the jumbo frame configuration of 9000 vs. 8992. There's that 8-byte difference, even though 802.2 networks can't carry jumbo frames, and IEEE has been otherwise derelict in their duty to standardize them. Alan Altmark Senior Managing z/VM and Linux Consultant Lab Services System z Delivery Practice IBM Systems & Technology Group ibm.com/systems/services/labservices office: 607.429.3323 mobile; 607.321.7556 alan_altm...@us.ibm.com IBM Endicott -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
OSA Gig supports jumbo..BUT routers/switches in the path must be set to do so as well. -Original Message- From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of Cohen, Sam Sent: Friday, January 29, 2016 9:36 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: MTU size Chris, If you look in the hardware documentation for the Open Systems Adapter (OSA), it has a "normal" MTU of 1492, so your Linux is matching the hardware settings. The OSA does support jumbo frames (8992) if your network is configured to use them. Thanks, Sam Cohen Levi, Ray & Shoup, Inc. -Original Message- From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of Christer Solskogen Sent: Friday, January 29, 2016 1:56 AM To: LINUX-390@VM.MARIST.EDU Subject: MTU size Hi! On all of my x86_64 Linux boxes the default MTU is 1500. But with SuSE (both 11 and 12) on s390 they seems to be 1492. How come? -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit https://urldefense.proofpoint.com/v2/url?u=http-3A__www.marist.edu_htbin_wlvindex-3FLINUX-2D390=CwIGaQ=JeEQba34pbypuH24NmkPeQmaR10xRxVB2BkB2ZkzOTA=GmMCqcCAb_PnEQRDX8l03QrVvBcsyg4O0tRLv9ePqZ8=7pIBRPm76XVqH4f98JJSm_vr9kFIvRMUEgPJdcONWpw=TMOMt7fTT8bx5zoPh2oyxY1GyNUjGGSWcXSPuqzN5VE= -- For more information on Linux on System z, visit https://urldefense.proofpoint.com/v2/url?u=http-3A__wiki.linuxvm.org_=CwIGaQ=JeEQba34pbypuH24NmkPeQmaR10xRxVB2BkB2ZkzOTA=GmMCqcCAb_PnEQRDX8l03QrVvBcsyg4O0tRLv9ePqZ8=7pIBRPm76XVqH4f98JJSm_vr9kFIvRMUEgPJdcONWpw=-PS0oG4SVARnKbpXXiiUKaiFA6NZe2djIP7wxxfDk5c=
Re: brtfs -u issue
>>> On 1/28/2016 at 10:26 AM, PHILIP TULLYwrote: > According to the documentation to change the uuid of a btrfs filesystem > in SLES 12 s390x, we should be using the btrfstune with the *u > option, but this doesn't seem to work. The result is we are unable to > mount a duplicate copy of a disk with BTRFS in two different mounts on > the same system. My understanding is that it is -U, not -u. Also, that it is not available in SLES12, but is in SLES12 SP1. Whether it actually works there or not, I cannot say. > Does anyone have any suggestions? Try SLES12 SP1? Not ideal, of course, but it might give you/us a data point. Mark Post -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/
Re: MTU size
Thanks everyone! I've learned something today as well! -- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 -- For more information on Linux on System z, visit http://wiki.linuxvm.org/