View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0004711||OXID eShop (all versions)||4.08. Cache||public||2012-11-16 14:21||2014-09-10 09:33|
|Product Version||4.7.1 / 5.0.1 revision 52468|
|Target Version||4.8.8 / 5.1.8||Fixed in Version||4.9.0_5.2.0_RC1|
|Summary||0004711: Trying to login with wrong credentials and varnish enabled, causes varnish to turn off|
|Description||When you try to login with fake logins, varnish turns off for some time. It starts working again without restarting after few minutes.|
|related to||0004687||resolved||aurimas.gladutis||User gets too many notification then makes a login with incorrect credentials|
|The problem is browser (session) dependent. When you try to login, varnish is shown inactive only in that browser. Other sessions is working correctly.|
After flushing all cache RP is show as not active for a few seconds.
After turning on RP for first time it also take some time to show RP as active.
varnishlog do not show those calls at all.
can you explain this please more precise?
>>varnishlog do not show those calls at all.
This would me the HTTP packets are not routed through varnish at all which is a known bug if you use pipe instead of pass
Reminder sent to: mark
Shop behaviour might be related to varnish suggestion to close connection after pipe https://www.varnish-cache.org/trac/wiki/VCLExamplePipe
Some shop clients told that this would improve Shop HIT count, as Varnish cache PIPE for IP address for several minutes. So Shop customer would go through PIPE after POST even if he make other request through GET.
However we could not reproduce this in our testing platform. Closing connection after PIPE did not decrease average response time.
We tried to use PASS instead of PIPE for POST request, however we got in to strange situation when NGINX returned 500 error.
>>as Varnish cache PIPE
no, this is because of Keep-Alive from apache behind. PIPE is just forwarding the HTTP packet as it is, which causes the apache2 to answer to the client directly. If keep-alive is actived the existing tcp connection is reused for other http request - going directly to apache/nginx and bypassing varnish. that's the reason why varnish suggests doing a connection close with pipe.
Using close can increase the HIT rate if you have pipes, but that's different in each customer setup.
Can I help you regarding the 500 errors? Can you provide details/logs?
|pipe changed to pass after post request|