-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256
All,
Maybe I can earn myself a beer.
>> On 3/6/20 13:44, Rainer Jung wrote:
>>> no, the status unfortunately is not available as an Apache env
>>> var.
>
>>> mod_proxy_ajp has a builtin provision for automatic env var
>>> forwarding: alle env vars named AJP_SOMETHING will be
>>> forwarded as request attribute SOMETHING. But I see no easy way
>>> of detecting drain mode and setting an env var and there seems
>>> to be nothing builtin. I remember having added the forwarding
>>> for the LB activation to mod_jk many years ago but well after
>>> mod_proxy_ajp was done by Mladen. It seems to be one of the
>>> missing features in mod_proxy_ajp.
>
>> I'd of course want this to be available in mod_proxy_balancer so
>> any protocol would get it.
>
>>> But of course it could be added ...
In Apache 2.4.x, mod_proxy_balancer.c looks like this starting at line
636:
/* Add session info to env. */
apr_table_setn(r->subprocess_env,
"BALANCER_SESSION_STICKY", sticky);
apr_table_setn(r->subprocess_env,
"BALANCER_SESSION_ROUTE", route);
This seems like a great place to add:
apr_table_setn(r->subprocess_env,
"BALANCER_WORKER_STATUS", ???);
The question is "what should the '???' be?"
The local "proxy_worker **worker" is the place to start. It looks like
I can get the status bits using:
(*worker)->s->status
So I could either convert that into a char* (because the environment
works on strings) or use ap_proxy_parse_wstatus to convert the status
into a human-readable char*.
In either case, what's the proper way to handle memory management for
strings in httpd/mod_proxy/etc.?
There is the brute-force yet safe:
char *status = apr_table_get(r->subprocess_env,
"BALANCER_WORKER_STATUS");
if(NULL == status) {
free(status);
}
status = (char*)malloc(5 * sizeof char); // 4 characters is enough + \0
if(NULL != status) {
if(0 <= snprintf(status, 5, "%d", (*worker)->s->status)) {
apr_table_setn(r->subprocess_env,
"BALANCER_WORKER_STATUS", status);
} else {
apr_table_unset(r->subprocess_env,
"BALANCER_WORKER_STATUS");
free(status);
}
} else {
apr_table_unset(r->subprocess_env,
"BALANCER_WORKER_STATUS");
}
This is a little wasteful, as it re-allocates a 5-byte buffer for
every request. I can get away with a /single/ allocation if I
understand the worker lifecycle a little better.
Is this moving toward the right approach?
Other thoughts:
* Would hex be a better representation?
* Would a spelled-out version be better? It looks like
ap_proxy_parse_wstatus will generate a string like "ONH".
- How to handle memory management of the return value?
(Is this what the apt_pool_r* is being passed-around for?)
Thanks,
- -chris
-----BEGIN PGP SIGNATURE-----
Comment: Using GnuPG with Thunderbird - https://www.enigmail.net/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=UElj
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
