Hi everyone, i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions. i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work? another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms? Thanks and best regards, -- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com>wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
i think drbd is more HA aware and works very well in HA setups. rsync is not recommended for hot/live mirroring and specially when doing it both directions. -m On Thu, Dec 30, 2010 at 9:27 PM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com>wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
I'm not sure I'd trust a rsync of a live database. On Thu, Dec 30, 2010 at 10:27 AM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com>wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- -Rupa
Can use Sqlite replication in between 2 instances by sqlite3_update_hook. Regds Sam On Thu, Dec 30, 2010 at 9:57 PM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com>wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
DRBD is a very interesting and very performing way to replicate a block device between two or more machine, usally in an active-passive way. And is very HA aware. In classic configuration, you have two machines that control each other state via heartbeat, and a floating IP address that is assigned to the "active" machine (in addition to the "real" IP address). The "active" machine is accessed from the HA application through the floating IP. The "passive" machine is there idle, but ready to take over in case the "active" goes down (for failure or maintenance). In that case the "passive" machine automatically becomes the "active", eg: gets the floating IP. The only drawback I can see in this setup (DRBD+nativesqlite instead of odbc) is that if the FS instance or the machine itself crashes very badly, maybe will trash the db, and because the sqlite db is replicated at filesystem level via DRDB to the other machine, you'll end up with a trashed sqlite db on the "new active" machine too. Having instead a logic SQL replica (mysql, postgres, whatever) between the two machines, spares you this possible point of failure. Also, if you use another heartbeat and another floating IP address dedicated to the odbc database access, you get HA database too (meaning: you can do an upgrade of the database engine on the "active" FS machine, because you moved only the floating IP of the dabase access to the other machine). -giovanni On Thu, Dec 30, 2010 at 6:07 PM, Sam <u2nsam@gmail.com> wrote:
Can use Sqlite replication in between 2 instances by sqlite3_update_hook.
Regds Sam
On Thu, Dec 30, 2010 at 9:57 PM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com> wrote:
Hi everyone, i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions. i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work? another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards, -- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Sincerely, Giovanni Maruzzelli Cell : +39-347-2665618
it needs to compile the kernel with drbd module... ----- Original Message ----- From: "Giovanni Maruzzelli" <gmaruzz@gmail.com> To: "FreeSWITCH Users Help" <freeswitch-users@lists.freeswitch.org> Sent: Thursday, December 30, 2010 1:00 PM Subject: Re: [Freeswitch-users] FreeSWITCH SBC Setup (HA mode) DRBD is a very interesting and very performing way to replicate a block device between two or more machine, usally in an active-passive way. And is very HA aware. In classic configuration, you have two machines that control each other state via heartbeat, and a floating IP address that is assigned to the "active" machine (in addition to the "real" IP address). The "active" machine is accessed from the HA application through the floating IP. The "passive" machine is there idle, but ready to take over in case the "active" goes down (for failure or maintenance). In that case the "passive" machine automatically becomes the "active", eg: gets the floating IP. The only drawback I can see in this setup (DRBD+nativesqlite instead of odbc) is that if the FS instance or the machine itself crashes very badly, maybe will trash the db, and because the sqlite db is replicated at filesystem level via DRDB to the other machine, you'll end up with a trashed sqlite db on the "new active" machine too. Having instead a logic SQL replica (mysql, postgres, whatever) between the two machines, spares you this possible point of failure. Also, if you use another heartbeat and another floating IP address dedicated to the odbc database access, you get HA database too (meaning: you can do an upgrade of the database engine on the "active" FS machine, because you moved only the floating IP of the dabase access to the other machine). -giovanni On Thu, Dec 30, 2010 at 6:07 PM, Sam <u2nsam@gmail.com> wrote:
Can use Sqlite replication in between 2 instances by sqlite3_update_hook.
Regds Sam
On Thu, Dec 30, 2010 at 9:57 PM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com> wrote:
Hi everyone, i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions. i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work? another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards, -- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Sincerely, Giovanni Maruzzelli Cell : +39-347-2665618 _______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
Linode has some great guides for setting it up: http://library.linode.com/linux-ha/ <http://library.linode.com/linux-ha/>I'm going with master-master mysql for now because I don't want the second machine to be completely passive, I'd like the ability to do something with it if I wished. -Avi On Thu, Dec 30, 2010 at 8:21 PM, Madovsky <infos@madovsky.org> wrote:
it needs to compile the kernel with drbd module...
----- Original Message ----- From: "Giovanni Maruzzelli" <gmaruzz@gmail.com> To: "FreeSWITCH Users Help" <freeswitch-users@lists.freeswitch.org> Sent: Thursday, December 30, 2010 1:00 PM Subject: Re: [Freeswitch-users] FreeSWITCH SBC Setup (HA mode)
DRBD is a very interesting and very performing way to replicate a block device between two or more machine, usally in an active-passive way. And is very HA aware.
In classic configuration, you have two machines that control each other state via heartbeat, and a floating IP address that is assigned to the "active" machine (in addition to the "real" IP address).
The "active" machine is accessed from the HA application through the floating IP.
The "passive" machine is there idle, but ready to take over in case the "active" goes down (for failure or maintenance). In that case the "passive" machine automatically becomes the "active", eg: gets the floating IP.
The only drawback I can see in this setup (DRBD+nativesqlite instead of odbc) is that if the FS instance or the machine itself crashes very badly, maybe will trash the db, and because the sqlite db is replicated at filesystem level via DRDB to the other machine, you'll end up with a trashed sqlite db on the "new active" machine too.
Having instead a logic SQL replica (mysql, postgres, whatever) between the two machines, spares you this possible point of failure.
Also, if you use another heartbeat and another floating IP address dedicated to the odbc database access, you get HA database too (meaning: you can do an upgrade of the database engine on the "active" FS machine, because you moved only the floating IP of the dabase access to the other machine).
-giovanni
On Thu, Dec 30, 2010 at 6:07 PM, Sam <u2nsam@gmail.com> wrote:
Can use Sqlite replication in between 2 instances by sqlite3_update_hook.
Regds Sam
On Thu, Dec 30, 2010 at 9:57 PM, Avi Marcus <Avi@amarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com> wrote:
Hi everyone, i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions. i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to
keep
.db files in sync between two fs boxes. is it likely to work? another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards, -- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE: http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Sincerely,
Giovanni Maruzzelli Cell : +39-347-2665618
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
That'll leave you wide open to having a corrupted db, drbd too. Personally i would use mysql cluster via odbc - all data on 2 nodes, which may or may not be on separate hardware to fs. Steve on iPhone On 30 Dec 2010, at 16:27, Avi Marcus <Avi@aMarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@gmail.com> wrote: Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
The CORE db for freeswitch doesn't really matter - it's all state information (other than voicemail DB), so if it really all goes to hell, e.g. in a crash, then you're no worse off than opening a clean instance of freeswitch. -Avi On Thu, Dec 30, 2010 at 7:44 PM, Steven Ayre <steveayre@gmail.com> wrote:
That'll leave you wide open to having a corrupted db, drbd too.
Personally i would use mysql cluster via odbc - all data on 2 nodes, which may or may not be on separate hardware to fs.
Steve on iPhone
On 30 Dec 2010, at 16:27, Avi Marcus <Avi@aMarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@<http://gmail.com> gmail.com> wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: <cyrenity@ekiga.net>cyrenity@ekiga.net mail: mustafa.pk@ <http://gmail.com>gmail.com web: <http://cyrenity.wordpress.com>cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list <FreeSWITCH-users@lists.freeswitch.org> FreeSWITCH-users@lists.freeswitch.org <http://lists.freeswitch.org/mailman/listinfo/freeswitch-users> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:<http://lists.freeswitch.org/mailman/options/freeswitch-users> http://lists.freeswitch.org/mailman/options/freeswitch-users <http://www.freeswitch.org>http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
The OP was about an HA implementation with live migration of calls from one machine to another without drop. The OPoster implemented this kind of HA correctly, and was asking about the possible drawbacks of changing the db storage from odbc (with a master-master active-passive replica, I suppose) to a native sqlite replicated via DRBD. I believe avoiding a clean restart is one of his requirement, or at least a strongly desired feature. -giovanni On 12/30/10, Avi Marcus <Avi@amarcus.com> wrote:
The CORE db for freeswitch doesn't really matter - it's all state information (other than voicemail DB), so if it really all goes to hell, e.g. in a crash, then you're no worse off than opening a clean instance of freeswitch. -Avi
On Thu, Dec 30, 2010 at 7:44 PM, Steven Ayre <steveayre@gmail.com> wrote:
That'll leave you wide open to having a corrupted db, drbd too.
Personally i would use mysql cluster via odbc - all data on 2 nodes, which may or may not be on separate hardware to fs.
Steve on iPhone
On 30 Dec 2010, at 16:27, Avi Marcus <Avi@aMarcus.com> wrote:
The other option for failover I was considering is the default sqlite - set up a job to rsync it to/from the slave constantly. Perhaps even throw that into a ramdisk, if that makes a difference (the kernal may be caching it anyway). -Avi
On Thu, Dec 30, 2010 at 2:14 PM, Ghulam Mustafa <mustafa.pk@<http://gmail.com> gmail.com> wrote:
Hi everyone,
i am trying to setup FreeSWITCH as a SBC in ha mode; for this purpose i have setup two FS boxes(identical) in HA mode using heartbeat, mysql as odbc backend and enabled track-calls in both sip profiles. Everything is working as expected i.e sip sessions successfully recovered during failover in both directions.
i am just curious and would like to know should i go for odbc or is it possible to avoid odbc and use drbd for /opt/switch/db directory to keep .db files in sync between two fs boxes. is it likely to work?
another question is related to performance when track-calls is enabled, we are supposed to handle ~300 sip concurrent calls around the clock in proxy_media mode, coming from our sip provider (on external profile) and forwarding to our call-center server (configured as internal sip gateway), we have plenty of good hardware (quad-core Xeon servers with 8Gigs ram), is it good to go :) or can raise alarms?
Thanks and best regards,
-- Ghulam Mustafa cell: +92 333.611.7681 sip: <cyrenity@ekiga.net>cyrenity@ekiga.net mail: mustafa.pk@ <http://gmail.com>gmail.com web: <http://cyrenity.wordpress.com>cyrenity.wordpress.com
_______________________________________________ FreeSWITCH-users mailing list <FreeSWITCH-users@lists.freeswitch.org> FreeSWITCH-users@lists.freeswitch.org <http://lists.freeswitch.org/mailman/listinfo/freeswitch-users> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:<http://lists.freeswitch.org/mailman/options/freeswitch-users> http://lists.freeswitch.org/mailman/options/freeswitch-users <http://www.freeswitch.org>http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
-- Sent from my mobile device Sincerely, Giovanni Maruzzelli Cell : +39-347-2665618
@Giovanni; Yes you are right, i have my desired setup in place. A 2 node mysql-cluster is in place, was just thinking about the possibility of sqlite db fs/block level replication. thank you everyone for putting thoughts on it. though my second question about track-calls performance went un-answered, i hope one of dev's can explain how it can degrade performance on a high load server. keeping scalability in view i'd love to know how a rpc or curl backend for core db will perform :) though i know it's not yet implemented neither there are any chances but just a thought. Regards, -- Ghulam Mustafa cell: +92 333.611.7681 sip: cyrenity@ekiga.net mail: mustafa.pk@gmail.com web: cyrenity.wordpress.com
participants (7)
-
Avi Marcus
-
Ghulam Mustafa
-
Giovanni Maruzzelli
-
Madovsky
-
Rupa Schomaker
-
Sam
-
Steven Ayre