cookie_monster Posted June 9, 2009 Author Posted June 9, 2009 What's the advantage of LACP mode over a card team other than the ability to use more than two connections?
spc-rocket Posted June 9, 2009 Posted June 9, 2009 What's the advantage of LACP mode over a card team other than the ability to use more than two connections? Its the same thing as teaming however with LACP both the switch and the remote note i.e. teamed server will load balance depending how much traffic is moving through it. There are some types of teaming i.e. ALB (adaptive load balancing which does the cleaver stuff at the server but the switch just treats it as normal 2 x 1Gb connection i.e. the switch is not doing load balancing or dynamically adjusting the traffic where as with LACP it does this dynamically and also other protocols are supported. Ash.
Ric_ Posted June 9, 2009 Posted June 9, 2009 @Duke: DO NOT drop to the Solaris shell - it is all logged and it will invalidate your warranty. @cookie_monster: The SP password is not necessarily the same as your BUI password - try logging in to the SP web interface (defaults are root/changeme). 2
cookie_monster Posted June 10, 2009 Author Posted June 10, 2009 Right it turns out that agregation wasn't configured correctly on the switch, a bit of fiddling in the CLI and it's sorted and I can log into the SP now as well Thanks.
kmount Posted June 10, 2009 Posted June 10, 2009 DO NOT drop to the Solaris shell - it is all logged and it will invalidate your warranty. Can't stress that enough; Sun mentioned a customer of theirs who'd had trouble which turned out to have stemmed from this and as Ric says it is all logged. Whilst on the subject of the kit, the Discovery Day was fantastic with loads of useful information on these beauties with a bright road map ahead. Be excited, I was
Duke Posted June 10, 2009 Posted June 10, 2009 Don't worry, I'm not even playing with the console let alone the shell! Just looked into it once when I ran into some problems during the initial setup. We ended up doing a factory reset from the BUI and all has been good since then.
nutso Posted June 10, 2009 Posted June 10, 2009 I had exactly the same problem as Butuz with the LACP aggregation. I ended up figuring out the destroy / create stuff on the command line to get it back up. The thing is that whenever I try selecting LACP aggregation and applying it, I get the message "The active destination is not part of the new network configuration." I went ahead anyway and the console said the link had the correct IP but I had no connectivity. Am I missing something with that error message, or is it more likely that the switch is giving me problems?
john Posted June 10, 2009 Posted June 10, 2009 Can't stress that enough; Sun mentioned a customer of theirs who'd had trouble which turned out to have stemmed from this and as Ric says it is all logged. Whilst on the subject of the kit, the Discovery Day was fantastic with loads of useful information on these beauties with a bright road map ahead. Be excited, I was I heard that story today as well as the discovery day (I was at the Manchester one). Very worthwhile, and informative. Great to also meet Phil an excellent and very knoledgable chap along with the rest of the team from Sun, so thanks to them all. And remember they know everything you do in the shell so behave!!!!
cookie_monster Posted June 10, 2009 Author Posted June 10, 2009 Currently my network config page looks like this: Interfaces: IPv4 static, 10.4.*.*/22, via aggr1 Datalinks: via nge0, nge1, nge2, nge3 but only two of the ports on the switch are part of the trunk, when I make the other two ports on the switch part of the trunk I loose the SAN connection. Where am I going wrong? I'm not quite sure i've got my head aroung the relationship between interfaces and datalinks. I'd like all four network connections to be seen as the same IP and connected to the switch via a LACP trunk.
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 Can someone familiar with HP switches suggest why LACP appears disabled in the status field, I’m pretty sure I’ve used the correct command to set the ports and if I unset it I loose the SAN. It appears to work but shows disabled in the LACP Status column but it also shows LACP in the type column. The command I issued was: 2910(config)# trunk ethernet 1-2 trk1 lacp Image of switch config here: http://www.buxton.derbyshire.sch.uk/grantpic/switch.jpg
Duke Posted June 11, 2009 Posted June 11, 2009 Question: To those of you who are making a single 4Gb trunk, do you find you really need that bandwidth and is your network really supplying enough data to maximise it? I would have thought you would have had bottlenecks elsewhere unless the SAN is just talking to servers that have a lot of NICs. Also, I remember reading somewhere that the BUI can put out quite a bit of traffic itself when you're doing a lot of analysis, would it not be better to have a 1Gb link dedicated for the BUI on one IP address that you know will always be available and a 3Gb link for data? Final question - assuming some of you are using the SAN for virtualisation, shouldn't you have a separate connection on the NIC for NFS/iSCSI/VMotion traffic on a different VLAN/IP range? Sorry for all the questions, just trying to figure out how I'm going to hook mine up. Chris
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 A good question about the link for VMotion will that require a dedicated link? I was going to have two connections trunked on a private network just to talk to the servers but I assumed that could be used for VMotion and so on. I might have the BUI on it's own IP as well at least for now as currently when I mess with the trunk I occasionally loose the SAN
Butuz Posted June 11, 2009 Posted June 11, 2009 (edited) Yes that's what I am trying to work out now - the best way of connecting it all. As far as the connection from the 7110 to the Core Switch - It will be a 4GB trunk on it's own "SAN" VLAN. Each ESX Server will then have a 2GB Trunk to the SAN Vlan for NFS, and a 2GB trunk out on to the Curriculum/Admin VLAN for VMs. Butuz Edited June 11, 2009 by Butuz
Butuz Posted June 11, 2009 Posted June 11, 2009 Can someone familiar with HP switches suggest why LACP appears disabled in the status field, My Switch does not show Trunk Status in the Web interface, I have to use a hyper terminal connection on mine to configure trunking. Your set up looks to be the same as mine though. I do not have flow control enabled on the trunk ports - don't know if that makes a difference? Butuz
Duke Posted June 11, 2009 Posted June 11, 2009 A good question about the link for VMotion will that require a dedicated link? I was going to have two connections trunked on a private network just to talk to the servers but I assumed that could be used for VMotion and so on. I think you'd be fine using the same private network that the servers are on, the main thing is to have your VMkernel traffic separate from your public network traffic, e.g. if you were using the SAN for CIFS direct to your users then make sure it's on a separate network, particularly as someone mentioned that VMotion copies the server memory across in plaintext. I'm thinking - 1x 1Gb for the BUI, 1x 1Gb for iSCSI/NFS to ESX, 2x 1Gb for CIFS to all our users (roaming profiles and documents, very large resources, etc). If I find that CIFS traffic is lower than I expected then I can give the virtualisation feed 2Gb instead. Anyone know how much an extra 4x 1Gb copper NIC costs? I assume it's just a generic PCIe card but that you need to buy it from Sun. Cheers, Chris
Butuz Posted June 11, 2009 Posted June 11, 2009 (edited) Deffo buy it from SUN. Have a word with Andy at Cutter maybe he can help with pricing. X7280A-2 Sun PCIe Dual Gigabit Ethernet UTP Low Profile Adapter, RoHS-6 Compliant £ 210.00 X4446A-Z Sun x4 PCIe Quad Gigabit Ethernet Low Profile Adapter for Rack Servers, RoHS-6 Compliant £ 440.00 I am also starting to think the in built 4 ports may not be enough! May be worth you looking at 10GBe Duke - would probably be alot better with regards to future proofing. Butuz Edited June 11, 2009 by Butuz 1
Duke Posted June 11, 2009 Posted June 11, 2009 Expensive but not unreasonable. My 7410 isn't even in production use and I haven't maxed out 1Gb yet so I'm probably planning too far ahead, but it's always good to have an idea of what I'm going to need long-term. Might as well see if I can get a discount on buying an extra NIC with the next 7410... EDIT: Don't tempt me with 10Gb! Chris
Soulfish Posted June 11, 2009 Posted June 11, 2009 I've ended up adding a quad port gigabit NIC to our 7110 so that I had more ports to play with to separate off for different tasks. Do make sure you get a Sun NIC though. After speaking to Phil it seems that in the worst case Sun can turn around and void your warranty for using third-party hardware. You do get one of those disposable anti-static bands however Was nice and easy to install as well. Simply stick it in the correct slot (there's guidelines for which slots should be filled depending on what your doing to spread load across different IO controllers etc), switch the box back on and it's detected and appears in the BUI straight away
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 (edited) From your diagram Butuz am I right in thinking that all of your SAN ports will be on a private network so you won't have any CIFS shares? Also i've read that jumbo frames are a must on ports to the SAN carrying traffic for your VM's but what about Flow Control should that be on as well? Edited June 11, 2009 by cookie_monster
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 This is driving me mad i've configured Card0 to be on my network so I can manage the box for now and this was working fine, I then configured Cards 2-4 into a trunk on a private range. After that I cannot contact the box untill I change my IP to the private range and connect to the trunked IP, I can however still ping the public IP. Both connections are configured to allow administration. Can anyone shed any light on where i'm configuring thjis wrong? Switch config http://www.buxton.derbyshire.sch.uk/grantpic/switch2.jpg SAN Card config http://www.buxton.derbyshire.sch.uk/grantpic/san.jpg
Butuz Posted June 11, 2009 Posted June 11, 2009 From your diagram Butuz am I right in thinking that all of your SAN ports will be on a private network so you won't have any CIFS shares? It was the plan, but have changed this afternoon and I have been trying to set up exactly what you have been setting up (Eth0 on network, Eth 1 - 3 on Trunked private network) and have had same problems as you - with the web interface dropping off as soon as I commit changes. I have been having to resort to the CLI to destroy the private network datalink and then hey presto the web interface pops back on to the curriculum network. Hrmm Butuz
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 I left mine in that same config for about 20 mins while I had lunch, when I came back I could open the web interface, I could ping it the whole time though, very odd. I'm going to update the switches and the SAN to the latest software.
pete Posted June 11, 2009 Posted June 11, 2009 Also i've read that jumbo frames are a must on ports to the SAN carrying traffic for your VM's but what about Flow Control should that be on as well? Let TCP/IP handle flow control if you can. More reading: Virtual Threads: Beware Ethernet flow control.
DMcCoy Posted June 11, 2009 Posted June 11, 2009 Can anyone shed any light on where i'm configuring thjis wrong? Switch config http://www.buxton.derbyshire.sch.uk/grantpic/switch2.jpg SAN Card config http://www.buxton.derbyshire.sch.uk/grantpic/san.jpg I don't think your LACP trunk is going to activate if you don't even have LACP enabled on those ports, in fact with LACP enabled you shouldn't have to assign ports to the trunk in the first place. You will need to telnet in to do any serious configuration or diagnostics on the procurve. The web interface is actually really rubbish compared to the rest of their functions.
cookie_monster Posted June 11, 2009 Author Posted June 11, 2009 (edited) @ DMcCoy, this is what's confusing me I have used the CLi to enable it but it still doesn't show in the interface. My connection to the SAN is now working so it must be ok right??? This is the command I used and it seemed to do the trick. 2910(config)# trunk ethernet 2-4 trk1 lacp http://www.buxton.derbyshire.sch.uk/grantpic/lacp.jpg Edited June 11, 2009 by cookie_monster
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now