SDP-223

C. Thomas Tyler
Closed
Fix EDGESERVER check in p4d_base regression.

Fix an issue introduced in @22215.  The change introduces a
check of the $EDGESERVER environment var in p4d_base, which
isn't reliably set when p4d_base is called.  Though p4d_base
now sources in backup_functions.sh, which has logic to set the
EDGESERVER variable in the set_vars() function, that set_vars()
function is never actually called.

We can't just call the set_vars() function, because it makes
'front door' checks (i.e. calls 'p4' commands) to determine if
the current instances is an edge server.  There are a few
problems here: 1) The p4d_base is an init script, and so can't
assume that front-door commands will work, since the server
might be down.  So the call to determine if it is an edge
server won't give the right result, and the 'p4login' call
will hang unnecessarily for a while.

Proposed fixes:
* Rewrite the logic in set_vers that sets EDGESERVER. It
should detect whether the current server is an edge server
in a way that runs fast and reliably even when 'p4d' is
down.  Do something like 'p4d -k db.<whatever> -jd -' and
find which database has server specs, and match that value
in th schema to the server spec (named $SERVERID), and check
that the value for the 'Services:' field of the server spec
has the code for 'edge-server.'

* Rewite the set_vars() so it only does innocuous things like
set variables (as the name implies), and doesn't try a 'p4 login'.
or anything else "active." Warnign: Need to make sure that doesn't
break anything, i.e. if any place calling set_vars() relies on
it doing 'p4login'.  This may need some refactoring and further
testing.
Status
Closed
Project
perforce-software-sdp
Severity
C
Reported By
tom_tyler
Reported Date
Modified By
perforce
Modified Date
Owned By
russell_jackson
Component
init
Type
Bug