Re: Appsession and persistance

Hi Erik,

Do you have any process restarting/reloading haproxy from time to time? You seem to have enough entries in your table.

Maybe you can send me your configuration in a separated mail so I can have a look :)

cheers

On Thu, Dec 15, 2011 at 6:57 PM, Erik Torlen <erik.torlen#apicasystem.com>wrote:

> Hi Baptiste,****
>
> ** **
>
> It gives me this:****
>
> ** **
>
> root#domU-12-31-39-0B-49-C1:/home/haproxy# echo "show table" | socat
> unix:/home/haproxy/status -****
>
> # table: api, type: string, size:102400, used:11846****
>
> ** **
>
> /E****
>
> ** **
>
> *From:* Baptiste [mailto:bedis9#gmail.com]
> *Sent:* den 15 december 2011 12:33
>
> *To:* Erik Torlen
> *Cc:* haproxy#formilux.org
> *Subject:* Re: Appsession and persistance****
>
> ** **
>
> Hi Erik****
>
> ** **
>
> Can you paste your stick-table definition?****
>
> You can get some information about your stick table using socat:****
>
> echo "show table" | socat unix:/path/to/haproxy.stats -****
>
> don't forget the last "-" which tells socat to print to stdout.****
>
> ** **
>
> a+****
>
> ** **
>
> On Thu, Dec 15, 2011 at 12:07 PM, Erik Torlen <erik.torlen#apicasystem.com>
> wrote:****
>
> Hi Baptiste,****
>
> ****
>
> I tried using stick-table and it worked better. But I still seeing
> sessions being ended because of the request being directed to the wrong
> backend server.****
>
> ****
>
> When I don’t have any persistence enabled I get around 50% errors. But
> using stick-tables give me around 15-20% errors. And the pattern of the
> errors looks pretty weird. It is almost like the sessions are released in
> some way after a certain time. (See the attached image which shows how the
> errors are appearing in groups every 20-30s or so during a 5min loadtest.)
> ****
>
> ****
>
> I tried increasing the len length and the size but the result was the same.
> ****
>
> ****
>
> ****
>
> ****
>
> Any ideas?****
>
> ****
>
> /E****
>
> ****
>
> ****
>
> -----Original Message-----
> From: Baptiste [mailto:bedis9#gmail.com]
> Sent: den 14 december 2011 17:25
> To: Erik Torlen
> Cc: haproxy#formilux.org
> Subject: Re: Appsession and persistance****
>
> ****
>
> Hi Erik,****
>
> ****
>
> Forget about appsession, it's weak :)****
>
> ****
>
> Use haproxy 1.5-dev7 and stick tables. In your backend, just add the****
>
> few line below:****
>
> stick-table type string len 32 size 10K****
>
> stick store-response set-cookie(PHPSESSID)****
>
> stick on url_param(session_id)****
>
> ****
>
> I recommand using with "peer" keyword to get multiple haproxy to sync****
>
> their tables.****
>
> So in case of reload of haproxy or in case of lb failure, you could****
>
> keep on redirecting users to the same backend.****
>
> ****
>
> cheers****
>
> ****
>
> ****
>
> ****
>
> On Wed, Dec 14, 2011 at 1:11 PM, Erik Torlen****
>
> <erik.torlen#apicasystem.com> wrote:****
>
> > Hi,****
>
> > ****
>
> > I am working with an application that does not use the cookie as a
> "normal" application would do.****
>
> > ****
>
> > Basically it receives PHPSESSID (currently two PHPSESSID is received,
> which I would try to solve with their devs) and the returns the sessid as a
> query string on every****
>
> > call it makes against the web application.****
>
> > ****
>
> > Example.****
>
> > Step 1****
>
> > ****
>
> > GET /login****
>
> > ...****
>
> > HTTP/1.1 200 OK****
>
> > ...****
>
> > Set-Cookie: PHPSESSID=m2b6m6rkp8l02lu5boflht2pq6; path=/****
>
> > ...****
>
> > ****
>
> > Step 2****
>
> > ****
>
> > GET /list/?session_id= m2b6m6rkp8l02lu5boflht2pq6****
>
> > ...****
>
> > ****
>
> > HTTP/1.1 200 OK****
>
> > ...****
>
> > ****
>
> > ****
>
> > I tried using appsession for this like:****
>
> > "appsession session_id len 27 timeout 1h request-learn mode query-string"
> ****
>
> > ****
>
> > But that does not seem to work.****
>
> > ****
>
> > I guess I need HAProxy to get the cookie value from the Set-Cookie
> header and then keeping track of it using the query-string.****
>
> > Is that possible to do? If so, how is it done?****
>
> > ****
>
> > /E****
>
> ** **
>