We key some alerts off of haproxy stats;csv and I'm trying to wrap my head
around a suspiciously low 5xx counter for a given server in a backend that
was spewing 503's and 500's in it's own log.
How does the 5xx counter for a given server get incremented in the case of
a backend connection reset? Is it possible that an http 200 flows by, but
an RST causes haproxy to *log* a 502 while *not* incrementing the 5xx
counter?
An example session ended in term state:
SH--
....which states that haproxy was waiting on headers and so I'm not sure how
my theory stands up given that the status code might have passed through or
not.
running haproxy 1.5.14
around a suspiciously low 5xx counter for a given server in a backend that
was spewing 503's and 500's in it's own log.
How does the 5xx counter for a given server get incremented in the case of
a backend connection reset? Is it possible that an http 200 flows by, but
an RST causes haproxy to *log* a 502 while *not* incrementing the 5xx
counter?
An example session ended in term state:
SH--
....which states that haproxy was waiting on headers and so I'm not sure how
my theory stands up given that the status code might have passed through or
not.
running haproxy 1.5.14