On 08/08/2013 04:14 PM, Hans Lellelid wrote:> > Sorry, I will give some better output tomorrow. I think when this > happened I was still using menu.c32 so I didn't see any output, it just > reloaded the menu. But I will respond back tomorrow after trying this > with more scientific rigour. I might be able to do the packet trace, > though that may be a little tricky to orchestrate since these are all > ESXi VMs. >Oh, that's important. VMwares PXE stack sucks big time. -hpa
On Thu, Aug 8, 2013 at 7:15 PM, H. Peter Anvin <hpa at zytor.com> wrote:> On 08/08/2013 04:14 PM, Hans Lellelid wrote: > > > > Sorry, I will give some better output tomorrow. I think when this > > happened I was still using menu.c32 so I didn't see any output, it just > > reloaded the menu. But I will respond back tomorrow after trying this > > with more scientific rigour. I might be able to do the packet trace, > > though that may be a little tricky to orchestrate since these are all > > ESXi VMs. > > > > Oh, that's important. VMwares PXE stack sucks big time. >That doesn't surprise me, but it's unfortunately what I have to work with. I'm sure I'll get something figured out. I figure worst-case I could compile custom gpxelinux.0 for each of the various (3) servers (i.e. with the gpxe script to forward to that URL) that I am trying to chainload from the first pxe server and try that out; I don't know if that stands a better chance of working, but I suppose if the problem is specific to pxechn.c32 it might? I freely admit to only having a basic understanding of what's going on here. Hans
On Thu, Aug 8, 2013 at 7:21 PM, Hans Lellelid <hans at velum.net> wrote:> On Thu, Aug 8, 2013 at 7:15 PM, H. Peter Anvin <hpa at zytor.com> wrote: > >> On 08/08/2013 04:14 PM, Hans Lellelid wrote: >> > >> > Sorry, I will give some better output tomorrow. I think when this >> > happened I was still using menu.c32 so I didn't see any output, it just >> > reloaded the menu. But I will respond back tomorrow after trying this >> > with more scientific rigour. I might be able to do the packet trace, >> > though that may be a little tricky to orchestrate since these are all >> > ESXi VMs. >> > >> >> Oh, that's important. VMwares PXE stack sucks big time. >> > > That doesn't surprise me, but it's unfortunately what I have to work with. > I'm sure I'll get something figured out. I figure worst-case I could > compile custom gpxelinux.0 for each of the various (3) servers (i.e. with > the gpxe script to forward to that URL) that I am trying to chainload from > the first pxe server and try that out; I don't know if that stands a better > chance of working, but I suppose if the problem is specific to pxechn.c32 > it might? I freely admit to only having a basic understanding of what's > going on here.1) I've actually had pretty decent luck with the VMware PXE stack on some platforms, including pushing 17+MB/s with another version of lpxelinux.0 and fetching 90+MiB initrd files. 2) I've generally has better luck loading pxelinux.0 first then gpxelinux.0. 3) Why not use the CONFIG directive/config.c32? modules that need to be at different versions? -- -Gene