Re: VM/ESA TCP/IP and VSWITCH
Well I'm guessing it can't. I have the following in my TCPIP PROFILE: DEVICE OSA1LCS600 ; LINKOSA1LNK ETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following in the TCPIP console log: DTCPRI385I Device OSA1: DTCPRI386I Type: LCS, Status: Not started DTCPRI387I Envelope queue size: 0 DTCPRI388I Address: 0600 I change it to this: DEVICE OSA1OSD600 ; LINKOSA1LNK QDIOETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following: DTCPRS007E Error encountered in reading PROFILE TCPIP *: DTCPAR123I Line 237: Invalid type field DTCPAR123I Line 239: LINK statement: Invalid type field DTCPAR123I Line 244: Unknown link name in HOME cmd DTCPAR123I Line 247: Unknown link name in GATEWAY cmd DTCPAR123I Line 252: Device not found Line 237 corresponds to my DEVICE statement and line 239 is the LINK statement. 600 - 602 is attached to my TCPIP on the VM/ESA 2.40 guest and I have NICDEF and GRANT for the VMEASA240 guest in z/VM Unless anyone can point out something obivious I'm going to chalk this up to experience. Thanks, Billy On 17 Mar 2010 at 10:13, The IBM z/VM Operating System wrote: Hello, Would VM/ESA running 2nd level under z/VM be able to use the VSWITCH? Thanks, Billy
Re: VM/ESA TCP/IP and VSWITCH
Hi Billy I'm no expert at this, but you've defined your device as a LCS device, not a QDIO. A VSwitch has to be QDIO, so you need to change your device (and link) statements accordingly. For instance, I have DEVICE OSAQDIO OSD 0440 PORTNAME OSA1 PRIROUTER AUTORESTART LINKETH0QDIOETHERNET OSAQDIO1 ane the linkname in the home statement points to ETH0 device 440 is a 'port' into the VSwitch. In your case, your TCPIP is in a 2nd level system, but it still has to be in QDIO mode. If your level of TCP/IP does not support QDIO, you will find out as soon as it tries to process the DEVICE staatement. Mike Hammock - Original Message - From: Billy Bingham To: IBMVM@LISTSERV.UARK.EDU Sent: Friday, March 19, 2010 12:17 PM Subject: Re: VM/ESA TCP/IP and VSWITCH Well I'm guessing it can't. I have the following in my TCPIP PROFILE: DEVICE OSA1LCS600 ; LINKOSA1LNK ETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following in the TCPIP console log: DTCPRI385I Device OSA1: DTCPRI386I Type: LCS, Status: Not started DTCPRI387I Envelope queue size: 0 DTCPRI388I Address: 0600 I change it to this: DEVICE OSA1OSD600 ; LINKOSA1LNK QDIOETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following: DTCPRS007E Error encountered in reading PROFILE TCPIP *: DTCPAR123I Line 237: Invalid type field DTCPAR123I Line 239: LINK statement: Invalid type field DTCPAR123I Line 244: Unknown link name in HOME cmd DTCPAR123I Line 247: Unknown link name in GATEWAY cmd DTCPAR123I Line 252: Device not found Line 237 corresponds to my DEVICE statement and line 239 is the LINK statement. 600 - 602 is attached to my TCPIP on the VM/ESA 2.40 guest and I have NICDEF and GRANT for the VMEASA240 guest in z/VM Unless anyone can point out something obivious I'm going to chalk this up to experience. Thanks, Billy On 17 Mar 2010 at 10:13, The IBM z/VM Operating System wrote: Hello, Would VM/ESA running 2nd level under z/VM be able to use the VSWITCH? Thanks, Billy
Re: VM/ESA TCP/IP and VSWITCH
The doc for TCP/IP for VM/ESA FL320 (which ran on VM/ESA 2.4) is still on line. There is no indication of QDIO device support in the Planning manual. Unless is was added after GA. On 03/19/2010 11:17 AM, Billy Bingham wrote: Well I'm guessing it can't. I have the following in my TCPIP PROFILE: DEVICE OSA1LCS600 ; LINKOSA1LNK ETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500 HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following in the TCPIP console log: DTCPRI385I Device OSA1: DTCPRI386I Type: LCS, Status: Not started DTCPRI387I Envelope queue size: 0 DTCPRI388I Address: 0600 I change it to this: DEVICE OSA1OSD600 ; LINKOSA1LNK QDIOETHERNET 0 OSA1 ; ARPAGE 5 ; HOME 10.1.1.93 OSA1LNK ; GATEWAY 10.1.1.1 =OSA1LNK 1500HOST DEFAULTNET=OSA1LNK 15000 ; TRANSLATE ; START OSA1 And I get the following: DTCPRS007E Error encountered in reading PROFILE TCPIP *: DTCPAR123I Line 237: Invalid type field DTCPAR123I Line 239: LINK statement: Invalid type field DTCPAR123I Line 244: Unknown link name in HOME cmd DTCPAR123I Line 247: Unknown link name in GATEWAY cmd DTCPAR123I Line 252: Device not found Line 237 corresponds to my DEVICE statement and line 239 is the LINK statement. 600 - 602 is attached to my TCPIP on the VM/ESA 2.40 guest and I have NICDEF and GRANT for the VMEASA240 guest in z/VM Unless anyone can point out something obivious I'm going to chalk this up to experience. Thanks, Billy -- Rich Smrcina Phone: 414-491-6001 http://www.linkedin.com/in/richsmrcina Catch the WAVV! http://www.wavv.org WAVV 2010 - Apr 9-13, 2010 Covington, KY
Re: VM/ESA TCP/IP and VSWITCH
To use a VSWITCH, the guest must support QDIO, and I'd be surprised if a that old VM/ESA would know about QDIO 2010/3/17 Billy Bingham billy.bingham...@suddenlink.net Hello, Would VM/ESA running 2nd level under z/VM be able to use the VSWITCH? Thanks, Billy -- Kris Buelens, IBM Belgium, VM customer support
Re: VM/ESA TCP/IP and VSWITCH
VM/ESA 2.4 had QDIO support (in fact it was the last release to have that module delivered with source). 2.3 might have had it, but I can't remember. I suspect it would work, but can't try it (no longer have a IPLable 2.4 system). From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Kris Buelens Sent: Thursday, March 18, 2010 1:22 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: VM/ESA TCP/IP and VSWITCH To use a VSWITCH, the guest must support QDIO, and I'd be surprised if a that old VM/ESA would know about QDIO
VM/ESA TCP/IP and VSWITCH
Hello, Would VM/ESA running 2nd level under z/VM be able to use the VSWITCH? Thanks, Billy
Re: TCP/IP and VSWITCH
On Thursday, 11/12/2009 at 02:13 EST, Frank M. Ramaekers framaek...@ailife.com wrote: I'll need a GATEWAY for the VSWITCH? Why? It can't get to the outside world, only other hosts in the VSWITCH. (I would think that any IP address in the network 10.1.20.128 would be sent to the VSWITCH adapter. Or am I misunderstanding you?) Without a route, the stack will take the shortest route and that will be the external connection. The GATEWAY lets you override that. Remember that in a multi-homed stack, it is a routing hop to move from one interface to another. You have: / VMip1---net1---VSEip1\ External net via OSA VMVSE \ VMip2---net2---VSEip2/ Internal net via VSWITCH So if you're on VM and reference VSEip1, the packet will exit VMip1 since it is on the same network as VSEip1. The GATEWAY entry forces the stack to take the extra hop *through* VSEip2 to get to VSEip1. Fair warning: If VSE is checking origin IP addresses for any reason, it will see VMip2. That brings up a good question, should something larger than 1500 be used for a VSWITCH? We are using a MTU size of 57344 (don't know where that came from). MTU sizes come from something the hardware. I recognize 57344 is the largest MTU allowed for a HiperSocket chpid that is defined with maximum frame size (MFS) of 64K (CHPARM=C0 in IOCP). For ethernet, the MTU comes from a setting in the [real] switch or adapter. Cisco supports jumbo frames that are up to 9K bytes, but OSA supports only up to 8992 bytes. Because the VSWITCH simulates OSA, the largest MTU is 8992. The choice of a larger or smaller MTU depends on where the traffic is going. If it's mostly between two adjacent hosts, larger is fine. If you're hopping through the adjacent host to somewhere else, then smaller is probably better. Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
I believe that in our case, most of the traffic (FTP) is external rather than between VMs. Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr. Fax: (254)741-5777 Waco, Texas 76701 From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Tom Huegel Sent: Wednesday, November 11, 2009 4:24 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH Frank, I had 7 VSE's that originally each had a dedicated OSA and changed all of them to use a single VSWITCH. I never saw a OSA capacity problem. In fact I saw some improvement, probably because a) all OSA ports went to the same network switch, and b) a fair amount of traffic was VSE to VSE, now that never hits the OSA ports, just the VSWITCH. Plus I gained the failover feature. Also I did not connect my VM TCPIP stack to the VSWITCH. On Wed, Nov 11, 2009 at 4:13 PM, Alan Altmark alan_altm...@us.ibm.com wrote: On Wednesday, 11/11/2009 at 04:31 EST, Frank M. Ramaekers framaek...@ailife.com wrote: That's great, if I was wanting to rework the entire mainframe network. My plans were just to route any intra-mainframe IP traffic onto a VSwitch and leave all of the external communication to the current method(s) (dedicated OSA). (You know the adage KISS). I do like the redundancy with VSWITCH with multiple OSAs though. (Maybe sometime in the future.) As a side note, did you discuss with your Network People first? To do what you want with VM TCP/IP means creation of another IP subnet and addresses and, possibly, the use of VIPA. That depends on whether or not you care about what IP address VM TCP/IP uses as an origin IP on outbound packets. Yes, reconfiguring network flows can be a non-trivial effort. That's why they deserve some thought before you deploy. Rule #1 of virtual networking: Never EVER make virtual network configuration changes without the express [written, preferably] approval of the Networking People. Just peeling off the packets to a particular host is easily done, but the ramifications of doing so are glued to the Law of Unintended Consequences. (What? I need VIPA just to do *that*? That means MPROUTE!) Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
Oh yes, I have complete control of 10.1.20.0/24. Within this I have only about 6 addresses assigned (and all 128). So, I decided to subnet it to 10.1.20.0/25, giving me two nets of 126 addresses. The 128 would be the internal (to the z9) addresses and the 128 external. Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, November 11, 2009 4:13 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Wednesday, 11/11/2009 at 04:31 EST, Frank M. Ramaekers framaek...@ailife.com wrote: That's great, if I was wanting to rework the entire mainframe network. My plans were just to route any intra-mainframe IP traffic onto a VSwitch and leave all of the external communication to the current method(s) (dedicated OSA). (You know the adage KISS). I do like the redundancy with VSWITCH with multiple OSAs though. (Maybe sometime in the future.) As a side note, did you discuss with your Network People first? To do what you want with VM TCP/IP means creation of another IP subnet and addresses and, possibly, the use of VIPA. That depends on whether or not you care about what IP address VM TCP/IP uses as an origin IP on outbound packets. Yes, reconfiguring network flows can be a non-trivial effort. That's why they deserve some thought before you deploy. Rule #1 of virtual networking: Never EVER make virtual network configuration changes without the express [written, preferably] approval of the Networking People. Just peeling off the packets to a particular host is easily done, but the ramifications of doing so are glued to the Law of Unintended Consequences. (What? I need VIPA just to do *that*? That means MPROUTE!) Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Thursday, 11/12/2009 at 08:34 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Oh yes, I have complete control of 10.1.20.0/24. Within this I have only about 6 addresses assigned (and all 128). So, I decided to subnet it to 10.1.20.0/25, giving me two nets of 126 addresses. The 128 would be the internal (to the z9) addresses and the 128 external. I've been Franked! is now added to my lexicon. Any *more* surprises? :-P The above does not match your configuration. You were showing subnet masks of 255.255.0.0, which precludes the subnetting you describe above. Oh, and you WILL need a GATEWAY statement in order to force the stack to take the indirect route: external IP HOST internal IP VSW0 0 (The MTU of zero causes it to use the MTU specified on the LINK VSW0 statement.) Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
Oh that was the default for the entire intranet, but since I'm slicing my little network up, it no longer applies. I'll need a GATEWAY for the VSWITCH? Why? It can't get to the outside world, only other hosts in the VSWITCH. (I would think that any IP address in the network 10.1.20.128 would be sent to the VSWITCH adapter. Or am I misunderstanding you?) That brings up a good question, should something larger than 1500 be used for a VSWITCH? We are using a MTU size of 57344 (don't know where that came from). Sorry for the suprises (they were unintended for the sake of brevity). Frank M. Ramaekers Jr. Systems Programmer; MCP, MCP+I, MCSE RHCE American Income Life Insurance Company Phone: (254) 761-6649 Fax: (254) 741-5777 From: The IBM z/VM Operating System on behalf of Alan Altmark Sent: Thu 11/12/2009 12:21 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Thursday, 11/12/2009 at 08:34 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Oh yes, I have complete control of 10.1.20.0/24. Within this I have only about 6 addresses assigned (and all 128). So, I decided to subnet it to 10.1.20.0/25, giving me two nets of 126 addresses. The 128 would be the internal (to the z9) addresses and the 128 external. I've been Franked! is now added to my lexicon. Any *more* surprises? :-P The above does not match your configuration. You were showing subnet masks of 255.255.0.0, which precludes the subnetting you describe above. Oh, and you WILL need a GATEWAY statement in order to force the stack to take the indirect route: external IP HOST internal IP VSW0 0 (The MTU of zero causes it to use the MTU specified on the LINK VSW0 statement.) Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
See below... Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Tuesday, November 10, 2009 4:46 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Tuesday, 11/10/2009 at 04:43 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Can someone help with the 'PROFILE TCPIP' settings for a VSWITCH that has only one other IP address on it (I'm in the testing stages. I have the DEVICE/LINK: DEVICE VSWA90 OSD A90 PORTNAME PORTA90 PORTNUMBER 0 NONROUTER AUTORESTART LINKVSW0 QDIOETHERNET VSWA90 I'm not sure about the HOME section, but I have: 10.1.20.20 255.255.0.0 ETH0 10.1.20.20 255.255.0.0 VSW0 I don't see a DEVICE and LINK for ETH0, so get rid of it. In any case, you can't assign the same IP address to two different interfaces. (Well, you might be able to but you better not.) Can't (I omitted them for brevity) it is the only way to everything else! And I have listed in the GATEWAY: 10.1.20.2 255.255.255.255 = VSW0 1500 Remove the above. You don't need it. A VSWITCH is a LAN, so you don't need (or want) point-to-point routes. 10.1.0.0 255.255.0.0 = ETH0 1500 Remove the above and code the MTU size on the LINK statement instead. DEFAULTNET 10.1.1.1 ETH0 1500 OK, assuming that 10.1.1.1 is Out There somewhere. Default route to everything else Basically, I do need, for now, a point to point. I can't subnet it, because of other devices on the same subnet (via the ETH0). What I need is a host route (no subnet)? --- The other host is a z/VSE system and it has: * Link to VSWITCH on VSWTMK00 DEFINE LINK,ID=AILVSW0,TYPE=OSAX,DEV=(A90,A91),DATAPATH=A92,- IPADDR=010.001.020.002 * - VSWITCH to z/VM TCP/IP stack DEFINE ROUTE,ID=AILVSWT,LINKID=AILVSW0,IPADDR=010.001.020.020 In other words: VSWITCH ETH0 10.1.20.2 --- 10.1.20.20 - Everything else z/VSEz/VM \ \ Everything else Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Wednesday, 11/11/2009 at 08:23 EST, Frank M. Ramaekers framaek...@ailife.com wrote: In other words: VSWITCH ETH0 10.1.20.2 --- 10.1.20.20 - Everything else z/VSEz/VM \ \ Everything else I am going to chastise myself (yet again) for dealing with a configuration problem without requiring the picture FIRST. Strange as it sounds, your picture is not valid or you aren't understanding VSWITCHes. Whenever a host touches a network of any sort (LAN or point-to-point), an IP address comes into existence. (Special dispensation given for unnumbered IP interfaces, which z/VM TCP/IP doesn't support.) If the two networks are not bridged by something outside of the hosts, then they are, by definition, separate networks. Every network gets its own subnet, even if if contains only two hosts. And within Subnet A, you never assign the IP addresses belonging to Subnet B. Never. In your drawing, I detect two networks: 1. A network labeled Everything else (10.1.0.0/16) 2. A network that has no label and no assigned subnet You cannot communicate on Network 2 until you assign it a subnet that is not within 10.1.0.0/16, and then assign IP addresses within that new subnet. If you simply trying to attach VM TCP/IP and z/VSE to a VSWITCH, the picture looks like 10.1.20.210.1.20.20 | ETH0| ETH0 (vNICs) --VSWITCH OSA | The VSWITCH is | an ethernet bridge ---Real network-- 10.1.0.0/16 They can talk to each other and out to the network. Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
On Wednesday, 11/11/2009 at 09:43 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Okay, that is what I'm use to (except for the z/VSE world). In the z/VSE world, you can have multiple interfaces with the same IP address, but subnetted down to the host (which z/VM doesn't appear to support): [snip] It appears that I cannot to this z/VM (i.e. cannot subnet down to a individual host) z/VM supports host routes, but they aren't necessary in this configuration. I have a vague feeling that you are trying to fake out the network, and that way lies both madness and damnation. If you are trying to define multiple interfaces for redundancy, you don't need them when using a VSWITCH. The VSWITCH handles the OSA failover transparently to the guests using the VSWITCH. So I'll put it this way: Tell us what you are trying to do (as opposed to 'configure') and then let's work on the best virtual network design to meet your needs. Given that design, we can build the configuration files to implement it. Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
Okay, I was originally trying to short circuit an IP route through the VSwitch (rather than creating a subnet and a new set of IP addresses). The z/VSE system would have only one address, 10.1.20.2 and z/VM would have a single address of 10.1.20.20. All the IP traffic would go to the ETH0 interface, except for those to this particular z/VSE (.2), which would use the VSWitch. We have done this with Hipersockets (or is it Hypersockets) from z/VSE to z/VSE. Now, I want to start moving these connections to VSwitches. So, on my first attempt, I was trying to the z/VSE TCP/IP to talk to our test z/VSE system via the VSwitch (keeping the same IP addresses). Does this help? Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, November 11, 2009 12:02 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Wednesday, 11/11/2009 at 09:43 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Okay, that is what I'm use to (except for the z/VSE world). In the z/VSE world, you can have multiple interfaces with the same IP address, but subnetted down to the host (which z/VM doesn't appear to support): [snip] It appears that I cannot to this z/VM (i.e. cannot subnet down to a individual host) z/VM supports host routes, but they aren't necessary in this configuration. I have a vague feeling that you are trying to fake out the network, and that way lies both madness and damnation. If you are trying to define multiple interfaces for redundancy, you don't need them when using a VSWITCH. The VSWITCH handles the OSA failover transparently to the guests using the VSWITCH. So I'll put it this way: Tell us what you are trying to do (as opposed to 'configure') and then let's work on the best virtual network design to meet your needs. Given that design, we can build the configuration files to implement it. Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Wednesday, 11/11/2009 at 01:58 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Okay, I was originally trying to short circuit an IP route through the VSwitch (rather than creating a subnet and a new set of IP addresses). The z/VSE system would have only one address, 10.1.20.2 and z/VM would have a single address of 10.1.20.20. All the IP traffic would go to the ETH0 interface, except for those to this particular z/VSE (.2), which would use the VSWitch. We have done this with Hipersockets (or is it Hypersockets) from z/VSE HiperSockets to z/VSE. Now, I want to start moving these connections to VSwitches. So, on my first attempt, I was trying to the z/VSE TCP/IP to talk to our test z/VSE system via the VSwitch (keeping the same IP addresses). The VSWITCH already has a built-in shortcut to other guests on the same [VLAN on] the same VSWITCH. You don't need any routing. When a guest sends packets out into the VSWITCH, CP will automatically short circuit them to other guests. It's a switch and that's what switches do - they look for local ports before they send it down the trunks to other switches. You can turn off the short circuit by using VSWITCH Isolation. Only in this case will outbound guest traffic be forced onto the OSA. And if you have a z10, the OSA will be told to force the packet onto the wire, bypassing the OSA LPAR-LPAR short circuit. Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
Oh, there is no OSA associated with the VSwitch (it's internal only). Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, November 11, 2009 1:40 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Wednesday, 11/11/2009 at 01:58 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Okay, I was originally trying to short circuit an IP route through the VSwitch (rather than creating a subnet and a new set of IP addresses). The z/VSE system would have only one address, 10.1.20.2 and z/VM would have a single address of 10.1.20.20. All the IP traffic would go to the ETH0 interface, except for those to this particular z/VSE (.2), which would use the VSWitch. We have done this with Hipersockets (or is it Hypersockets) from z/VSE HiperSockets to z/VSE. Now, I want to start moving these connections to VSwitches. So, on my first attempt, I was trying to the z/VSE TCP/IP to talk to our test z/VSE system via the VSwitch (keeping the same IP addresses). The VSWITCH already has a built-in shortcut to other guests on the same [VLAN on] the same VSWITCH. You don't need any routing. When a guest sends packets out into the VSWITCH, CP will automatically short circuit them to other guests. It's a switch and that's what switches do - they look for local ports before they send it down the trunks to other switches. You can turn off the short circuit by using VSWITCH Isolation. Only in this case will outbound guest traffic be forced onto the OSA. And if you have a z10, the OSA will be told to force the packet onto the wire, bypassing the OSA LPAR-LPAR short circuit. Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Wednesday, 11/11/2009 at 03:25 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Oh, there is no OSA associated with the VSwitch (it's internal only). Ah. New Network = New Subnet = New IP Addresses. There is special support in Linux and z/OS, and maybe the z/VSE stacks?, provide a HiperSocket Accelerator which may be enabled when you put the same IP addy on different interfaces. Dunno. z/VM TCP/IP doesn't have that capability. But you're making this way too hard, Frank. :-) Just take the OSA(s) away from z/VSE and VM TCP/IP and give it/them to the VSWITCH. No worries about new subnets, no host routes, no duplicate IP addresses. And you inherit OSA failover capability without requiring any kind of dynamic routing daemons. Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
I can, but we have about 8 OSAs (and associated z/VSEs). Combining them into one VSWITCH with a single adapter to the outside world might be a problem (bandwidth to the outside). Can you have more than one OSA associated with the VSwitch? Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, November 11, 2009 2:37 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Wednesday, 11/11/2009 at 03:25 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Oh, there is no OSA associated with the VSwitch (it's internal only). Ah. New Network = New Subnet = New IP Addresses. There is special support in Linux and z/OS, and maybe the z/VSE stacks?, provide a HiperSocket Accelerator which may be enabled when you put the same IP addy on different interfaces. Dunno. z/VM TCP/IP doesn't have that capability. But you're making this way too hard, Frank. :-) Just take the OSA(s) away from z/VSE and VM TCP/IP and give it/them to the VSWITCH. No worries about new subnets, no host routes, no duplicate IP addresses. And you inherit OSA failover capability without requiring any kind of dynamic routing daemons. Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Wednesday, 11/11/2009 at 03:44 EST, Frank M. Ramaekers framaek...@ailife.com wrote: I can, but we have about 8 OSAs (and associated z/VSEs). Combining them into one VSWITCH with a single adapter to the outside world might be a problem (bandwidth to the outside). Can you have more than one OSA associated with the VSwitch? Do you have any other surprises? :-) So you have one OSA chpid per z/VSE? To get bandwidth 1 OSA, you have to run with Link Aggregation which requires capability in the physical switch. You could choose instead to create 8 VSWITCHes, each with one OSA or, more likely, 4 VSWITCHes, each with 2 OSAs (1 per for failover). Alan Altmark z/VM Development IBM Endicott
Re: TCP/IP and VSWITCH
That's great, if I was wanting to rework the entire mainframe network. My plans were just to route any intra-mainframe IP traffic onto a VSwitch and leave all of the external communication to the current method(s) (dedicated OSA). (You know the adage KISS). I do like the redundancy with VSWITCH with multiple OSAs though. (Maybe sometime in the future.) Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 -Original Message- From: The IBM z/VM Operating System [mailto:ib...@listserv.uark.edu] On Behalf Of Alan Altmark Sent: Wednesday, November 11, 2009 3:05 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: TCP/IP and VSWITCH On Wednesday, 11/11/2009 at 03:44 EST, Frank M. Ramaekers framaek...@ailife.com wrote: I can, but we have about 8 OSAs (and associated z/VSEs). Combining them into one VSWITCH with a single adapter to the outside world might be a problem (bandwidth to the outside). Can you have more than one OSA associated with the VSwitch? Do you have any other surprises? :-) So you have one OSA chpid per z/VSE? To get bandwidth 1 OSA, you have to run with Link Aggregation which requires capability in the physical switch. You could choose instead to create 8 VSWITCHes, each with one OSA or, more likely, 4 VSWITCHes, each with 2 OSAs (1 per for failover). Alan Altmark z/VM Development IBM Endicott _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
Frank, I had 7 VSE's that originally each had a dedicated OSA and changed all of them to use a single VSWITCH. I never saw a OSA capacity problem. In fact I saw some improvement, probably because a) all OSA ports went to the same network switch, and b) a fair amount of traffic was VSE to VSE, now that never hits the OSA ports, just the VSWITCH. Plus I gained the failover feature. Also I did not connect my VM TCPIP stack to the VSWITCH. On Wed, Nov 11, 2009 at 4:13 PM, Alan Altmark alan_altm...@us.ibm.comwrote: On Wednesday, 11/11/2009 at 04:31 EST, Frank M. Ramaekers framaek...@ailife.com wrote: That's great, if I was wanting to rework the entire mainframe network. My plans were just to route any intra-mainframe IP traffic onto a VSwitch and leave all of the external communication to the current method(s) (dedicated OSA). (You know the adage KISS). I do like the redundancy with VSWITCH with multiple OSAs though. (Maybe sometime in the future.) As a side note, did you discuss with your Network People first? To do what you want with VM TCP/IP means creation of another IP subnet and addresses and, possibly, the use of VIPA. That depends on whether or not you care about what IP address VM TCP/IP uses as an origin IP on outbound packets. Yes, reconfiguring network flows can be a non-trivial effort. That's why they deserve some thought before you deploy. Rule #1 of virtual networking: Never EVER make virtual network configuration changes without the express [written, preferably] approval of the Networking People. Just peeling off the packets to a particular host is easily done, but the ramifications of doing so are glued to the Law of Unintended Consequences. (What? I need VIPA just to do *that*? That means MPROUTE!) Alan Altmark z/VM Development IBM Endicott
TCP/IP and VSWITCH
Can someone help with the 'PROFILE TCPIP' settings for a VSWITCH that has only one other IP address on it (I'm in the testing stages. I have the DEVICE/LINK: DEVICE VSWA90 OSD A90 PORTNAME PORTA90 PORTNUMBER 0 NONROUTER AUTORESTART LINKVSW0 QDIOETHERNET VSWA90 I'm not sure about the HOME section, but I have: 10.1.20.20 255.255.0.0 ETH0 10.1.20.20 255.255.0.0 VSW0 And I have listed in the GATEWAY: 10.1.20.2 255.255.255.255 = VSW0 1500 10.1.0.0 255.255.0.0 = ETH0 1500 DEFAULTNET 10.1.1.1 ETH0 1500 The other host I have on the VSWITCH (now) is 10.1.20.2 Thanks, Frank M. Ramaekers Jr. Systems Programmer MCP, MCP+I, MCSE RHCE American Income Life Insurance Co. Phone: (254)761-6649 1200 Wooded Acres Dr.Fax: (254)741-5777 Waco, Texas 76710 _ This message contains information which is privileged and confidential and is solely for the use of the intended recipient. If you are not the intended recipient, be aware that any review, disclosure, copying, distribution, or use of the contents of this message is strictly prohibited. If you have received this in error, please destroy it immediately and notify us at privacy...@ailife.com.
Re: TCP/IP and VSWITCH
On Tuesday, 11/10/2009 at 04:43 EST, Frank M. Ramaekers framaek...@ailife.com wrote: Can someone help with the 'PROFILE TCPIP' settings for a VSWITCH that has only one other IP address on it (I'm in the testing stages. I have the DEVICE/LINK: DEVICE VSWA90 OSD A90 PORTNAME PORTA90 PORTNUMBER 0 NONROUTER AUTORESTART LINKVSW0 QDIOETHERNET VSWA90 I'm not sure about the HOME section, but I have: 10.1.20.20 255.255.0.0 ETH0 10.1.20.20 255.255.0.0 VSW0 I don't see a DEVICE and LINK for ETH0, so get rid of it. In any case, you can't assign the same IP address to two different interfaces. (Well, you might be able to but you better not.) And I have listed in the GATEWAY: 10.1.20.2 255.255.255.255 = VSW0 1500 Remove the above. You don't need it. A VSWITCH is a LAN, so you don't need (or want) point-to-point routes. 10.1.0.0 255.255.0.0 = ETH0 1500 Remove the above and code the MTU size on the LINK statement instead. DEFAULTNET 10.1.1.1 ETH0 1500 OK, assuming that 10.1.1.1 is Out There somewhere. Alan Altmark z/VM Development IBM Endicott
