[c-nsp] IOS 12.2.33 SRA in 7600

[Mike Butash]

Just out of curiosity, what compelling reasons did you have (or how many 
beers) to put SRA on a box in production?  ;)

I'd looked through the features for SRA, there were only a few that 
really looked to be worth investigating, but for the most part it seemed 
not worth the time given I knew I'd be experiencing issues such as you 
describe as well.  It never fails just how many issues one finds when 
actually poking at a 6500 in real world situations these days, 
especially in 12.2SX, now apparently SR too.  I really feel like I'm 
being used as cheap Q/A for Cisco sometimes...

-mb


Richard A Steenbergen wrote:
> On Fri, Sep 08, 2006 at 06:22:57PM +0700, Ibrahim wrote:
>> Hi,
>>
>> Anybody have deployed IOS 12.2.33 SRA/1  in operation router 7600  ? Any
>> major bugs ?
> 
> Mixed bag of experience here. Hit a couple of major bugs that were 
> complete showstoppers early on. A lot of documented ones, some 
> undocumented ones, my personal favorite was doing a clear ip bgp (hard) on 
> some bgp neighbors, and having the tcp sessions drop on the remote end, 
> but having the SRA side continue to see them up and live. No amount of 
> clearing or shut/no shutting would make them reestablish or convince the 
> SRA box that they were actually down, had to reload to clear it. But, 
> couldn't manage to reproduce that later on, and SRA deployments so far 
> have been otherwise relatively uneventful (being careful not to 
> "unnecessarily poke" the box of course). Basically if you find that deep 
> dark spot in the back of your mind called "7500 mode", aka the mode of 
> thinking where you force yourself to not look at the box funny or 
> otherwise taunt it to fail, it generally won't fail. :)
> 
> I'll give you another example of a really funny issue I hit yesterday... 
> SRA box, trying to upgrade to SRA1 via ftp, and the image file was being 
> corrupted on download. Filesize as listed in dir correct, but the md5 hash 
> on the file was not, and it looked like it wasn't writing the last little 
> bit of the file correctly. Further, it was being written down in such a 
> way that when you did copy disk0:thenewimage disk0:anotherfile, the md5 
> have on the second copy was also completely different than the original or 
> the first copy, and so on. After a few hours of head scratching, 
> formatting, CF replacing, etc, the problem was finally traced to FTP under 
> SRA. By using tftp, or ftp running from older SXF code, the problem went 
> away completely.
> 
> Summary is: SRA1 is a good step in the right direction, and if you have 
> the right testicular fortitude it can be made to work quite successfully, 
> but it clearly has more than a few outstanding issues left to be resolved. 
> In fact, as I write this, I just got a traceback in the CLI as the result 
> of doing a traceroute. Cute. :)
>