So now I looked for more of those "SC"'s in the log, from our monitoring
and it appeared first around 13:38:01.
Around 13:37:54 a reload was issued by puppet or rundeck.
So right now, it seems that something happened during the reload which
affected UNIX sockets.
On 2020-03-27 15:00, Christian Ruppert wrote:
Hi Olivier,
On 2020-03-27 14:50, Olivier Houchard wrote:
Hi Christian,
On Fri, Mar 27, 2020 at 02:37:41PM +0100, Christian Ruppert wrote:
Hi list,
we have some weird issues now, the second time, that *some* SSL
sockets
seem to be broken as well as stats sockets.
HTTP seems to work fine, still, SSL ones are broken however. It
happened
at least on 2.1.3 and *perhaps* on 2.1.2 as well. We're not sure
whether
the first time was on 2.1.2 or 2.1.3.
The one that failed today was updated yesterday, so HAProxy has an
uptime of about 24h.
We're using threads. default + HTTP is using 1 thread, 1 is dedicated
for a TCP listener/Layer-4, one is for RSA only and all the rest is
for
ECC.
[...]
The problem ocurred arount 13:40 (CET, in case it matters at some
point)
Any ideas so far?
So basically, it used to work, and suddenly you get errors on any TLS
connection ?
Yeah, right now it looks like that way.
If you still have the TCP stat socket working, can you show the output
of "show fd" ?
Oh, it's the http stats listener that's still working. Not sure
whether it accepts any commands to be honest.
pid = 21313 (process #1, nbproc = 1, nbthread = 8)
uptime = 0d 1h56m48s
system limits: memmax = unlimited; ulimit-n = 1574819
maxsock = 1574819; maxconn = 786432; maxpipes = 0
current conns = 6; current pipes = 0/0; conn rate = 43/sec; bit rate =
219.704 kbps
Running tasks: 1/1158; idle = 100 %
Thanks !
Olivier
--
Regards,
Christian Ruppert
