Hi,
Finally decided to get memcached up and running with Roundcube -- awesome work there.
I ran into a problem getting it to go at first though. Turned on all the logging in memcache and found this relevant bit:
<28 get d5f4fbf9a78ea7c63f2919f1ba544fa0
FOUND KEY d5f4fbf9a78ea7c63f2919f1ba544fa0 28 sending key d5f4fbf9a78ea7c63f2919f1ba544fa0 28 END
28: going from conn_parse_cmd to conn_mwrite 28: going from conn_mwrite to conn_new_cmd 28: going from conn_new_cmd to conn_waiting 28: going from conn_waiting to conn_read 28: going from conn_read to conn_parse_cmd <28 delete d5f4fbf9a78ea7c63f2919f1ba544fa0 1
28 CLIENT_ERROR bad command line format. Usage: delete <key> [noreply]
...which eventually led me to http://www.php.net/manual/en/memcache.delete.php where this problem is mentioned in comments. The PHP docs there specifically say not to include the timeout parameter when calling memcache->delete(), but unfortunately, due to a bug in the current php5-memcached package in Debian, not including the parameter causes php to automatically append a "1" to the end of the delete command, which breaks memcached.
In rcube_session.php, I mucked about with the first line of mc_destroy(), changing it to:
$ret = $this->memcache->delete($key, 0);
...which makes everything happy, despite being incorrect.
I searched the dev- and users- mailing list archives and didn't find any other references to this particular problem, so I'm just mentioning it here in case anybody else gets bitten by it.
Thanks,
- R.
Rob Sheldon wrote:
Hi,
Finally decided to get memcached up and running with Roundcube -- awesome work there.
I ran into a problem getting it to go at first though. Turned on all the logging in memcache and found this relevant bit:
[...]
In rcube_session.php, I mucked about with the first line of mc_destroy(), changing it to:
$ret = $this->memcache->delete($key, 0);...which makes everything happy, despite being incorrect.
I don't have a problem adding this to your code and becoming "incorrect" unless anybody else reports problems with that change. PHP usually doesn't complain about additional arguments passed to a function call, only if there are too few arguments passed.
~Thomas _______________________________________________ List info: http://lists.roundcube.net/dev/ BT/aba52c80
On Sat, Jan 7, 2012 at 2:11 PM, Thomas Bruederli roundcube@gmail.comwrote:
Rob Sheldon wrote:
Hi,
Finally decided to get memcached up and running with Roundcube -- awesome work there.
I ran into a problem getting it to go at first though. Turned on all the logging in memcache and found this relevant bit:
[...]
In rcube_session.php, I mucked about with the first line of mc_destroy(), changing it to:
$ret = $this->memcache->delete($key, 0);...which makes everything happy, despite being incorrect.
I don't have a problem adding this to your code and becoming "incorrect" unless anybody else reports problems with that change. PHP usually doesn't complain about additional arguments passed to a function call, only if there are too few arguments passed.
+1 for the change: http://docs.php.net/manual/en/memcache.delete.php
This states that it's not recommend to use the timeout paramter, but using 0 is safe and shouldn't break anything anywhere else.
Till
List info: http://lists.roundcube.net/dev/ BT/aba52c80
On Jan 07 2012 08:50 am, till wrote:
On Sat, Jan 7, 2012 at 2:11
PM, Thomas Bruederli <roundcube@gmail.com [1]> wrote:
Rob Sheldon
wrote:
Hi,
Finally decided to get memcached up and
running with Roundcube --
awesome work there.
I ran into
a problem getting it to go at first though. Turned on all
the
logging in memcache and found this relevant bit:
[...]
In rcube_session.php, I mucked about with the first line of
mc_destroy(), changing it to:
$ret =
$this->memcache->delete($key, 0);
...which makes everything
happy, despite being incorrect.
I don't have a problem adding
this to your code and becoming "incorrect"
unless anybody else
reports problems with that change. PHP usually doesn't
complain about
additional arguments passed to a function call, only if
there are too
few arguments passed.
+1 for the change:
http://docs.php.net/manual/en/memcache.delete.php [2]
This states
that it's not recommend to use the timeout paramter, but using 0 is safe and shouldn't break anything anywhere else.
Till
...assuming that the PHP dev team doesn't change their mind later, which, y'know, never happens...
I also had to make a similar change in /program/include/rcube_cache.php:
private function delete_record($key, $index=true)
{
if ($this->type == 'memcache')
$this->db->delete($this->ckey($key), 0);
...otherwise, the first time you sign out, you can't sign back in again. Looks like Roundcube tries to delete a previous cache id during login.
Those are the only two changes I've needed to make and everything has worked great so far. I try not to alter Roundcube's code at all (makes future upgrades a little less easy) but memcache was just too tempting.
Thanks,
- R.
On Sat, Jan 7, 2012 at 6:54 PM, Rob Sheldon rob@associatedtechs.com wrote:
**
On Jan 07 2012 08:50 am, till wrote:
On Sat, Jan 7, 2012 at 2:11 PM, Thomas Bruederli roundcube@gmail.comwrote:
Rob Sheldon wrote:
Hi,
Finally decided to get memcached up and running with Roundcube -- awesome work there.
I ran into a problem getting it to go at first though. Turned on all the logging in memcache and found this relevant bit:
[...]
In rcube_session.php, I mucked about with the first line of mc_destroy(), changing it to:
$ret = $this->memcache->delete($key, 0);...which makes everything happy, despite being incorrect.
I don't have a problem adding this to your code and becoming "incorrect" unless anybody else reports problems with that change. PHP usually doesn't complain about additional arguments passed to a function call, only if there are too few arguments passed.
+1 for the change: http://docs.php.net/manual/en/memcache.delete.php
This states that it's not recommend to use the timeout paramter, but using 0 is safe and shouldn't break anything anywhere else. Till
...assuming that the PHP dev team doesn't change their mind later, which, y'know, never happens...
Memcache is not a core extension but since a lot of contributors to it are familiar with (or neck-deep in) the PHP process, follows BC principles throughout. If you have any issues, I can put you in touch with people who work on it. I happen to know a few of them. Just let me know. ;)
I also had to make a similar change in /program/include/rcube_cache.php:
private function delete_record($key, $index=true) { if ($this->type == 'memcache') $this->db->delete($this->ckey($key), 0);...otherwise, the first time you sign out, you can't sign back in again. Looks like Roundcube tries to delete a previous cache id during login.
Those are the only two changes I've needed to make and everything has worked great so far. I try not to alter Roundcube's code at all (makes future upgrades a little less easy) but memcache was just too tempting.
Thanks,
- R.
+1 for that change as well.
Till
List info: http://lists.roundcube.net/dev/ BT/aba52c80
-
Rob Sheldon -
Thomas Bruederli -
till