There is no point to use "old" numbers in recent commit.
Also avoid number duplication. The messages are the same but in different code path, so having different numbers makes sense.
This also avoids a warning when running:
make update-log-msg-tags
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1861333 13f79535-47bb-0310-9956-ffa450edef68
balancer_process_balancer_worker() to create/enable/disable
workers from another modules via a provider.
No functional changes in this first commit.
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1859235 13f79535-47bb-0310-9956-ffa450edef68
Otherwise, when BalancerPersist is enabled, we might fail (re)starting with a
saved slotmem if a BalancerMember was renamed (e.g. PR 62308's port change).
This is because the renamed member is considered a new one, while the old name
still reserves a slot, so we end up missing one room. The overall number of
members does not change so the sizes check succeeds thus the persisted slotmem
is not invalidated. Yet the slomem is not really compatible.
By clearing inuse flags, we still allow for slots to be reused if their index
did not change (thanks to ap_proxy_find_balancershm() and slotmem_fgrab() at
the given index), but will also invalidate renamed slots which don't match the
index. We now have the correct behaviour by slot and the server (re)starts in
any case.
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1831938 13f79535-47bb-0310-9956-ffa450edef68
r1831394 is the right follow to r1830800 to preserve "inherited" slotmems in
children processes.
While at it, comment on the expectations from mod_proxy_balancer w.r.t.
slotmem_attach() implementations, and eventually how we could improve the API
later (w/o backporting to 2.4.x).
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1831396 13f79535-47bb-0310-9956-ffa450edef68
Don't call slotmem_attach() if the slot is already initialized/reused, that
used to work previously because the returned error status is ignored, but
r1830800 changed the function to forcibly reset the returned slot pointer to
NULL first.
There is no point to call slotmem_attach() in this case, the slot is already
initialized because it's either inherited/fork()ed on Unixes, or on Windows
the attachment happened in post_config/slotmem_create() already.
One case where slotmem_attach() would fail is when balancers are defined
globally (main server) and some virtualhosts don't have/use any mod_proxy
configuration/directive. In this case their proxy server config is a pointer
copy of the main server thus their slots are already initialized in both
post_config and child_init hooks. While the case was already handled in
the balancer_post_config(), it did not in balancer_child_init() where we
relied on slotmem_attach() to be a noop.
[Reverted by r1831396]
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1831218 13f79535-47bb-0310-9956-ffa450edef68
used as drop-in replacements for unusable workers in the same load balancer set. This differs
from hot standbys which are only used when all workers in a set are unusable. PR 61140.
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1828890 13f79535-47bb-0310-9956-ffa450edef68
Rework server_rec ID so that it doesn't change on restart (or stop/start)
unless it's Host(s)/IP(s):port(s), ServerName and/or ServerAlias(es) changed.
The goal being to reuse SHMs (and persisted files) names as much as possible,
with minimal bindings to configuration changes (as far as mod_proxy_balancer
is concerned). So if the ServerName and first Host/IP:port are unique we use
that first, otherwise the ServerAlias(es) and other Host(s)/IP(s):port(s) are
also taken into account, and finally if that's still not enough the server
index is also used (pathological case handled for correctness with regard to
the underlying mod_slotmem_shm's reuse code).
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1823564 13f79535-47bb-0310-9956-ffa450edef68
restart when new ones are added. PR 58024.
Since slotmem_create() issues a strict check on the size of an existing
slot before reusing it, it won't reuse existing balancers and members when
new ones are added during restart (whereas growth margins would allow it).
Fix this by using slotmem_attach() first and if it succeeds do the checks
based on the returned previous number of existing entries.
[Reverted by r1698123]
git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/trunk@1696960 13f79535-47bb-0310-9956-ffa450edef68
