This has me thinking... we should likely do something to better error-check the store/restore aspects of slotmem. Even some sort of quick checksum would be better than what we have now. :/
Gotta mull this over a bit more. On Nov 8, 2013, at 1:29 PM, Jim Jagielski <[email protected]> wrote: > Mostly due to the "persist" feature... Right now it just does > a memcpy (basically). If people don't use it then it's not > an issue. > > > On Nov 8, 2013, at 10:00 AM, Thomas Eckert <[email protected]> > wrote: > >>> I'd be +1 in adjusting all of those fields bigger, but I'm guessing >>> that constitutes an API change for proxy... >> >> API change, why is that ? At least the shm size stuff doesn't look like >> having a lot of implications other then memory consumption - to make sure of >> this assumption is why I asked in the first place. I can well imagine >> setting those defaults at build time, possibly with a directive for >> performance fine tuning at post-build time. >> >> >> >> On Fri, Nov 8, 2013 at 3:35 PM, Jim Jagielski <[email protected]> wrote: >> Yeah, it's basically for performance and storage reasons (since those >> strings are stored in shm)... Nowadays I don't think shm is such an >> expensive commodity, though I can imagine some setups where the >> default sizes allowed by the kernel could be kinda small. >> >> I'd be +1 in adjusting all of those fields bigger, but I'm guessing >> that constitutes an API change for proxy... >> >> On Nov 8, 2013, at 5:17 AM, Thomas Eckert <[email protected]> >> wrote: >> >>> I'm looking at an issue with this log message >>> >>> AH00526: Syntax error on line 6 of myconfig.conf: BalancerMember worker >>> hostname (aaaaaaaa-bbbb-cccc-dd-eee-ffffffffff.us-east-1.elb.amazonaws.com) >>> too long >>> >>> with the root cause being (modules/proxy/mod_proxy.h) >>> >>> #define PROXY_WORKER_MAX_HOSTNAME_SIZE 64 >>> >>> and I'm wondering why this is. I did not see any comments on why there is >>> this limit, so I assume it's for performance reasons. >>> >>> Also, I'm wondering what the implications (if any) of altering this limit >>> are. >> >> >
