I posted originally my issue on askubuntu but I think this will be a
In simple terms
I am having issues with https handshakes. I am currently using nginx but
is most likely not an nginx issue.
Web clients such as browsers will sometimes present “SSL connection
Apache benchmark will spit out several error lines and will report
1-10% failures. Errors below will appear in random order but the first
is more common.
(1) Benchmarking mysite.net (be patient)…SSL read failed (1) - closing
failed or bad record mac:s3_pkt.c:486:
(2) SSL read failed (1) - closing connection
128494120003296:error:140943FC:SSL routines:SSL3_READ_BYTES:sslv3 alert
record mac:s3_pkt.c:1262:SSL alert number 20
Ubuntu 12.04 64bit with all updates and patches installed, server
OpenSSL dynamically linked:
which nginx | grep ssl
libssl.so.1.0.0 => /lib/x86_64-linux-gnu/libssl.so.1.0.0
strings /lib/x86_64-linux-gnu/libssl.so.1.0.0 | grep "^OpenSSL "
OpenSSL 1.0.1 14 Mar 2012
Nginx server config (with limited cyphers)
1.0.1 14 Mar 2012
#dpkg -s libssl1.0.0
So far, the only workaround I found, is to narrow down available
Instead of using Mozilla Intermediate set, I found these would work
Second option is to downgrade to stock nginx (1.1.19-1ubuntu0.7)
#Things I tried
Because I am mainly using nginx as a proxy / load balancer, I tried
replacing nginx with HA-Proxy 1.5. Unfortunately I got the same problem.
I tried compiling nginx 1.6.3 with openssl 1.0.1m - no change.
On-line https/ssl validity tester did not found any issues with any
Disabling other nginx sites did not help either.
#Things I noticed
Interestingly this problem does not occur when using apache
from the server itself or it’s immediate neighbours, but it does happen
connecting from outside of the data centre. Apparently DC guys (coreix)
claim not to have any DDOS prevention system in front of the servers
would cause such an issue.
This issue is happening mainly on one of the https domains and is
sporadic for remaining two - hosted on the same box.
Posted at Nginx Forum: