Re: Kannel 1.5.0 Debian packages
Hi,
[ Sorry for X-post ]
On Thu, 2011-03-24 at 10:26, Alan McNatty wrote:
Hi Alex,
Thanks for rolling us a new deb.
Thanks also for building with Postgres support - that will save me
rolling my own :)
Just a couple of questions:
1) I notice we have the usual binaries ... but I thought there was talk
of run_kannel_box being deprecated since we have --parachute --daemonize
capability. What are people doing these days by default - see attached
init.d script for example.
I built it to be as close as possible to official Debian package because
I don't want to 'surprise' users/admins during upgrade from stable.
2) (this is likely a longer term / roadmap question) Any chance sqlbox
(and possibly others, e.g smppbox) could be included as a binary as
well?
sqlbox (0.7.2 version) is shipped with Debian stable. I have made sqlbox
deb from SVN package just for test but didn't posted it because I don't
know how it works because I never tried to use or play with it.
It is sounded like more and more folk are using these as standard and it
would be advantageous to the project to have them included in standard
deb.
I built Opensmppbox deb packages few months age for me to play with it
but didn't posted it to Alex because I'm not sure how SVN version is
good to put on Kannel site for download.
I realise they are not in the standard build so possibly this is a
question more around when the source / build trees might be merged (i.e
--enable-sqlbox, etc)? So likely some work involved here - and best of
all likely involves autoconf, etc compiler magic that troubles many of
us.
Cheers,
Alan
p.s sorry for the cross-post .. just reply-all'ing
On Wed, 2011-03-23 at 12:20 +0100, Alexander Malysh wrote:
Hi All,
just uploaded debs for 1.5.0
to https://redmine.kannel.org/projects/kannel/files
please check and give me/Milan feedback.
Thanks,
Alexander Malysh
Am 22.03.2011 um 16:24 schrieb Milan P. Stanic:
Hi Alex,
On Tue, 2011-03-22 at 14:37, Alexander Malysh wrote:
Hi Milan,
as far as I see, configuration should be fine. Could you please
upload debs and diff tar to some place
and we will push it to download section then.
I will upload it later and send the URL to you in private mail.
Thanks,
Alexander Malysh
Am 22.03.2011 um 14:09 schrieb Milan P. Stanic:
Hi Alex,
On Tue, 2011-03-22 at 09:56, Alexander Malysh wrote:
Hi Milan,
that would be great to have debian packages for 1.5.0
I forgot to mention in original post that I sent these packages
to
someone in Romania who is tried it on Ubuntu. He told me that
the same
binary packages which I run on Debian also works well on Ubuntu
but I
forget which release he runs.
could you please answer some questions:
- how these packages compiled
I patched 1.5.0 source with official Debian diff (actually tar
file)
because my intention is that the packages I build shouldn't
differ a lot
from Debian shipped packages. It is because of that if someone
upgrade
it from official Debian release to newly built packages s/he
doesn't
have to change anything in configuration files.
- which DB support is enabled
Sqlite3, Postgresql and MySQL.
- compiler/configure flags used
Here is the relevant section from debian/rules files:
./configure \
--host=$(DEB_HOST_GNU_TYPE) \
--build=$(DEB_BUILD_GNU_TYPE) \
--enable-warnings \
--prefix=/usr \
--mandir=\$${prefix}/share/man \
--infodir=\$${prefix}/share/info \
--enable-docs --enable-pam --enable-pcre \
--enable-ssl --with-ssl=/usr \
--with-mysql --with-mysql-dir=/usr \
--with-sqlite3 \
--with-pgsql --with-pgsql-dir=/usr
And here is relevant section from Makefile:
LIBS=-lpq -lmysqlclient_r -lssl -ldl -lpam -lpcreposix -lrt
-lresolv
-lnsl -lm -lpthread -lxml2 -L/usr/lib -lpcreposix -lpcre
-L/usr/lib
-lcrypto -lssl -rdynamic -L/usr/lib/mysql -lmysqlclient_r
-L/usr/local/lib -lsqlite3 -L/usr/lib
CFLAGS=-D_REENTRANT=1 -I. -Igw -g -O2 -D_XOPEN_SOURCE=600
-D_BSD_SOURCE
-D_LARGE_FILES= -I/usr/include/libxml2 -Wall
-Wmissing-prototypes
-Wmissing-declarations -Wnested-externs -Winline -Wformat
-Wformat-security -Wmissing-format-attribute -I/usr/include
-I/usr/include/mysql -I/usr/include/postgresql
LDFLAGS= -rdynamic
Do you have some place to upload packages and we will push
those then to download section?
Yes, I can upload it but would like to know first if the
configuration
above is correct. If not, I can rebuild it with your (or someone
else)
suggestions.
Thanks,
Alexander Malysh
DLRs for concatenated messages
Hi all, I'm seeing some strangeness with DLR handling of concatenated messages sent via SMPP in Kannel 1.5.0. I'm using MySQL for DLR storage and only the first message of the set generates a DLR in the database. My provider, however, follows the specs and returns a DLR for each part. Kannel recognizes the first DLR and updates the database. When the DLRs for the remaining parts arrive, Kannel tries to find them in the database and fails, spitting out an error (see below). Specifically: 1. When the submit_sm_resp PDU for the first message part arrives, Kannel happily creates the DLR in the database: 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:37 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:37 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:37 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:37 [5106] [10] DEBUG: sequence_number: 17 = 0x0011 2011-03-25 02:03:37 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:37 [5106] [10] DEBUG:Octet string at 0xc8cbc50: 2011-03-25 02:03:37 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:37 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:37 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 36 36 36 31 38 33 1300989817666183 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:37 [5106] [10] DEBUG:Octet string dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: DLR[mysql]: Adding DLR smsc=MY-SMPP-CONNECTION, ts=13009898176661830402, src=test, dst=9, mask=31, boxc=smsbox 2011-03-25 02:03:37 [5106] [10] DEBUG: adding DLR entry into database 2011-03-25 02:03:37 [5106] [10] DEBUG: sql: INSERT INTO `smsgw_kannel_sms_dlr` (`smsc`, `timestamp`, `source`, `destination`, `service`, `url`, `mask`, `boxc_id`, `status`) VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0) 2. When the submit_sm_resp PDU for the subsequent message parts arrive, Kannel doesn't create a database DLR but instead appears to create one in memory (note the missing SQL): 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:38 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:38 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:38 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:38 [5106] [10] DEBUG: sequence_number: 18 = 0x0012 2011-03-25 02:03:38 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:38 [5106] [10] DEBUG:Octet string at 0xc8d1d80: 2011-03-25 02:03:38 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:38 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:38 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 39 33 34 31 38 34 1300989817934184 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:38 [5106] [10] DEBUG:Octet string dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: creating DLR message 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: DLR = 3. When the deliver_sm PDU arrives for the first message part, it's correctly matched against the database, processed and deleted: 2011-03-25 02:03:48 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:48 [5106] [10] DEBUG: type_name: deliver_sm 2011-03-25 02:03:48 [5106] [10] DEBUG: command_id: 5 = 0x0005 2011-03-25 02:03:48 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: sequence_number: 8 = 0x0008 2011-03-25 02:03:48 [5106] [10] DEBUG: service_type: NULL 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr_ton: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr_npi: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr: 9 2011-03-25 02:03:48 [5106] [10] DEBUG: dest_addr_ton: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: dest_addr_npi: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: destination_addr: test 2011-03-25 02:03:48 [5106] [10] DEBUG: esm_class: 4 = 0x0004 2011-03-25 02:03:48 [5106] [10] DEBUG: protocol_id: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: priority_flag: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: schedule_delivery_time: NULL 2011-03-25 02:03:48 [5106] [10] DEBUG: validity_period: NULL 2011-03-25 02:03:48 [5106] [10] DEBUG: registered_delivery: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: replace_if_present_flag: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: data_coding: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: sm_default_msg_id: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: sm_length: 165 = 0x00a5 2011-03-25 02:03:48 [5106] [10] DEBUG: short_message:
Re: DLRs for concatenated messages
Hi That is the way kannel handles dlr for concatenated messages. There has been several threads regarding how to improve this but nothing has been done. |-| Envíe y Reciba Datos y mensajes de Texto (SMS) hacia y desde cualquier celular y Nextel en el Perú, México y en mas de 180 paises. Use aplicaciones 2 vias via SMS y GPRS online Visitenos en www.perusms.NET www.smsglobal.com.mx y www.pravcom.com On Thu, Mar 24, 2011 at 1:29 PM, Toby Phipps toby.phi...@nexmedia.com.sg wrote: Hi all, I'm seeing some strangeness with DLR handling of concatenated messages sent via SMPP in Kannel 1.5.0. I'm using MySQL for DLR storage and only the first message of the set generates a DLR in the database. My provider, however, follows the specs and returns a DLR for each part. Kannel recognizes the first DLR and updates the database. When the DLRs for the remaining parts arrive, Kannel tries to find them in the database and fails, spitting out an error (see below). Specifically: 1. When the submit_sm_resp PDU for the first message part arrives, Kannel happily creates the DLR in the database: 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:37 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:37 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:37 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:37 [5106] [10] DEBUG: sequence_number: 17 = 0x0011 2011-03-25 02:03:37 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string at 0xc8cbc50: 2011-03-25 02:03:37 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:37 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:37 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 36 36 36 31 38 33 1300989817666183 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: DLR[mysql]: Adding DLR smsc=MY-SMPP-CONNECTION, ts=13009898176661830402, src=test, dst=9, mask=31, boxc=smsbox 2011-03-25 02:03:37 [5106] [10] DEBUG: adding DLR entry into database 2011-03-25 02:03:37 [5106] [10] DEBUG: sql: INSERT INTO `smsgw_kannel_sms_dlr` (`smsc`, `timestamp`, `source`, `destination`, `service`, `url`, `mask`, `boxc_id`, `status`) VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0) 2. When the submit_sm_resp PDU for the subsequent message parts arrive, Kannel doesn't create a database DLR but instead appears to create one in memory (note the missing SQL): 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:38 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:38 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:38 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:38 [5106] [10] DEBUG: sequence_number: 18 = 0x0012 2011-03-25 02:03:38 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:38 [5106] [10] DEBUG: Octet string at 0xc8d1d80: 2011-03-25 02:03:38 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:38 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:38 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 39 33 34 31 38 34 1300989817934184 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:38 [5106] [10] DEBUG: Octet string dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: creating DLR message 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: DLR = 3. When the deliver_sm PDU arrives for the first message part, it's correctly matched against the database, processed and deleted: 2011-03-25 02:03:48 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:48 [5106] [10] DEBUG: type_name: deliver_sm 2011-03-25 02:03:48 [5106] [10] DEBUG: command_id: 5 = 0x0005 2011-03-25 02:03:48 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: sequence_number: 8 = 0x0008 2011-03-25 02:03:48 [5106] [10] DEBUG: service_type: NULL 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr_ton: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr_npi: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: source_addr: 9 2011-03-25 02:03:48 [5106] [10] DEBUG: dest_addr_ton: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: dest_addr_npi: 0 = 0x 2011-03-25 02:03:48 [5106] [10] DEBUG: destination_addr: test 2011-03-25 02:03:48 [5106] [10] DEBUG: esm_class: 4 = 0x0004 2011-03-25
RE: DLRs for concatenated messages
Hi Alvaro, Thanks for the feedback. Any insight into why it was implemented this way? i.e. is there reason why each of the submit_sm_resp PDUs shouldn't generate a separate DLR in the DB? The only reason I can come up with offhand is that the current way probably handles misbehaving SMPP servers better - servers that send one DLR per concatenated message rather than per part. However, the current (broken, IMHO) behaviour could easily be maintained with a config parameter. Before I scope a patch to address this, any comments or disagreements? Also, I haven't tried this with memory-only DLRs - is this an issue limited to SQL DLR storage? From looking at my logs, it seems that a DLR is indeed created in memory for the second and subsequent message parts, just not written to the DB. Thanks, Toby -Original Message- From: Alvaro Cornejo [mailto:cornejo.alv...@gmail.com] Sent: Friday, March 25, 2011 1:47 AM To: Toby Phipps Cc: users@kannel.org Subject: Re: DLRs for concatenated messages Hi That is the way kannel handles dlr for concatenated messages. There has been several threads regarding how to improve this but nothing has been done. |--- --| Envíe y Reciba Datos y mensajes de Texto (SMS) hacia y desde cualquier celular y Nextel en el Perú, México y en mas de 180 paises. Use aplicaciones 2 vias via SMS y GPRS online Visitenos en www.perusms.NET www.smsglobal.com.mx y www.pravcom.com On Thu, Mar 24, 2011 at 1:29 PM, Toby Phipps toby.phi...@nexmedia.com.sg wrote: Hi all, I'm seeing some strangeness with DLR handling of concatenated messages sent via SMPP in Kannel 1.5.0. I'm using MySQL for DLR storage and only the first message of the set generates a DLR in the database. My provider, however, follows the specs and returns a DLR for each part. Kannel recognizes the first DLR and updates the database. When the DLRs for the remaining parts arrive, Kannel tries to find them in the database and fails, spitting out an error (see below). Specifically: 1. When the submit_sm_resp PDU for the first message part arrives, Kannel happily creates the DLR in the database: 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:37 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:37 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:37 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:37 [5106] [10] DEBUG: sequence_number: 17 = 0x0011 2011-03-25 02:03:37 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string at 0xc8cbc50: 2011-03-25 02:03:37 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:37 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:37 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 36 36 36 31 38 33 1300989817666183 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: DLR[mysql]: Adding DLR smsc=MY-SMPP-CONNECTION, ts=13009898176661830402, src=test, dst=9, mask=31, boxc=smsbox 2011-03-25 02:03:37 [5106] [10] DEBUG: adding DLR entry into database 2011-03-25 02:03:37 [5106] [10] DEBUG: sql: INSERT INTO `smsgw_kannel_sms_dlr` (`smsc`, `timestamp`, `source`, `destination`, `service`, `url`, `mask`, `boxc_id`, `status`) VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0) 2. When the submit_sm_resp PDU for the subsequent message parts arrive, Kannel doesn't create a database DLR but instead appears to create one in memory (note the missing SQL): 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:38 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:38 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:38 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:38 [5106] [10] DEBUG: sequence_number: 18 = 0x0012 2011-03-25 02:03:38 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:38 [5106] [10] DEBUG: Octet string at 0xc8d1d80: 2011-03-25 02:03:38 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:38 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:38 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 39 33 34 31 38 34 1300989817934184 2011-03-25 02:03:38 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:38 [5106] [10] DEBUG: Octet string dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: creating DLR message 2011-03-25 02:03:38 [5106] [10] DEBUG: SMSC[MY-SMPP-CONNECTION]: DLR = 3.
NoSQL and Kannel
Hi all, Has anyone put any thought into implementing Kannel DLR storage (and the basics of sqlbox MT injection) in a NoSQL datastore? We are using Redis (http://redis.io) for other projects and are very happy with it. Its key-based storage/retrieval and strong support for arrays as atomic queues would make it perfect for Kannel, especially given the transient nature of this data and Redis' memory speed with disk backing. The main advantages in my mind are that during heavy loads disk access would be minimized, DB polling would disappear (to be replaced with blocking array pops) and the SQL overhead would be removed replaced with set/fetch by keys. No particular reason this would need to be Redis - memcached, MongoDB, Cassandra etc. would also probably work well assuming they have the array/queue support. Any thoughts? It wouldn't be a small project, but basing the code on the existing implementations of SQL DLRs and sqlbox would mean a big jump start. Toby.
Re: NoSQL and Kannel
I've been playing with kannel and memcached a little bit, but nothing worth posting so far. The unreliable nature of memcached (because of its unreplicated horizontal scaling) doesn't make it very suitable for DLR's in fact (e.g., a node going down and zapping all the dlrs stored in it). Other noSQL engines would make more sense. Regards, Alex On Thu, Mar 24, 2011 at 8:12 PM, Toby Phipps toby.phi...@nexmedia.com.sgwrote: Hi all, Has anyone put any thought into implementing Kannel DLR storage (and the basics of sqlbox MT injection) in a NoSQL datastore? We are using Redis ( http://redis.io) for other projects and are very happy with it. Its key-based storage/retrieval and strong support for arrays as atomic queues would make it perfect for Kannel, especially given the transient nature of this data and Redis’ memory speed with disk backing. The main advantages in my mind are that during heavy loads disk access would be minimized, DB polling would disappear (to be replaced with blocking array pops) and the SQL overhead would be removed replaced with set/fetch by keys. No particular reason this would need to be Redis – memcached, MongoDB, Cassandra etc. would also probably work well assuming they have the array/queue support. Any thoughts? It wouldn’t be a small project, but basing the code on the existing implementations of SQL DLRs and sqlbox would mean a big jump start… Toby.
Re: DLRs for concatenated messages
Hi From my understanting - I'm not a developper- the issue is what to do when only one part of the concatenated fails -resend the whole/failed part- and how/when to validate the original full initial message. it seems that type of messafe handling will require to have a separate queue/buffer for concatenated messages in order to be able to track the whole message. Maybe you can do a search on the mailing list for concatenated messages and will find the threads between devs people. Regards Alvaro |-| Envíe y Reciba Datos y mensajes de Texto (SMS) hacia y desde cualquier celular y Nextel en el Perú, México y en mas de 180 paises. Use aplicaciones 2 vias via SMS y GPRS online Visitenos en www.perusms.NET www.smsglobal.com.mx y www.pravcom.com On Thu, Mar 24, 2011 at 2:00 PM, Toby Phipps toby.phi...@nexmedia.com.sg wrote: Hi Alvaro, Thanks for the feedback. Any insight into why it was implemented this way? i.e. is there reason why each of the submit_sm_resp PDUs shouldn't generate a separate DLR in the DB? The only reason I can come up with offhand is that the current way probably handles misbehaving SMPP servers better - servers that send one DLR per concatenated message rather than per part. However, the current (broken, IMHO) behaviour could easily be maintained with a config parameter. Before I scope a patch to address this, any comments or disagreements? Also, I haven't tried this with memory-only DLRs - is this an issue limited to SQL DLR storage? From looking at my logs, it seems that a DLR is indeed created in memory for the second and subsequent message parts, just not written to the DB. Thanks, Toby -Original Message- From: Alvaro Cornejo [mailto:cornejo.alv...@gmail.com] Sent: Friday, March 25, 2011 1:47 AM To: Toby Phipps Cc: users@kannel.org Subject: Re: DLRs for concatenated messages Hi That is the way kannel handles dlr for concatenated messages. There has been several threads regarding how to improve this but nothing has been done. |--- --| Envíe y Reciba Datos y mensajes de Texto (SMS) hacia y desde cualquier celular y Nextel en el Perú, México y en mas de 180 paises. Use aplicaciones 2 vias via SMS y GPRS online Visitenos en www.perusms.NET www.smsglobal.com.mx y www.pravcom.com On Thu, Mar 24, 2011 at 1:29 PM, Toby Phipps toby.phi...@nexmedia.com.sg wrote: Hi all, I'm seeing some strangeness with DLR handling of concatenated messages sent via SMPP in Kannel 1.5.0. I'm using MySQL for DLR storage and only the first message of the set generates a DLR in the database. My provider, however, follows the specs and returns a DLR for each part. Kannel recognizes the first DLR and updates the database. When the DLRs for the remaining parts arrive, Kannel tries to find them in the database and fails, spitting out an error (see below). Specifically: 1. When the submit_sm_resp PDU for the first message part arrives, Kannel happily creates the DLR in the database: 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:37 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:37 [5106] [10] DEBUG: command_id: 2147483652 = 0x8004 2011-03-25 02:03:37 [5106] [10] DEBUG: command_status: 0 = 0x 2011-03-25 02:03:37 [5106] [10] DEBUG: sequence_number: 17 = 0x0011 2011-03-25 02:03:37 [5106] [10] DEBUG: message_id: 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string at 0xc8cbc50: 2011-03-25 02:03:37 [5106] [10] DEBUG: len: 20 2011-03-25 02:03:37 [5106] [10] DEBUG: size: 21 2011-03-25 02:03:37 [5106] [10] DEBUG: immutable: 0 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 31 33 30 30 39 38 39 38 31 37 36 36 36 31 38 33 1300989817666183 2011-03-25 02:03:37 [5106] [10] DEBUG: data: 30 34 30 32 0402 2011-03-25 02:03:37 [5106] [10] DEBUG: Octet string dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: SMPP PDU dump ends. 2011-03-25 02:03:37 [5106] [10] DEBUG: DLR[mysql]: Adding DLR smsc=MY-SMPP-CONNECTION, ts=13009898176661830402, src=test, dst=9, mask=31, boxc=smsbox 2011-03-25 02:03:37 [5106] [10] DEBUG: adding DLR entry into database 2011-03-25 02:03:37 [5106] [10] DEBUG: sql: INSERT INTO `smsgw_kannel_sms_dlr` (`smsc`, `timestamp`, `source`, `destination`, `service`, `url`, `mask`, `boxc_id`, `status`) VALUES (?, ?, ?, ?, ?, ?, ?, ?, 0) 2. When the submit_sm_resp PDU for the subsequent message parts arrive, Kannel doesn't create a database DLR but instead appears to create one in memory (note the missing SQL): 2011-03-25 02:03:38 [5106] [10] DEBUG: SMPP PDU 0xc8ad960 dump: 2011-03-25 02:03:38 [5106] [10] DEBUG: type_name: submit_sm_resp 2011-03-25 02:03:38
Re: NoSQL and Kannel
I have written DLR storage adapters for Redis, Memcache and Tokyo Tyrant, but these were for consulting and are not open source. I will check to see if any of the customers will allow them to be published. However, they were fairly trivial to implement using the gw/dlr_* sources as a template. Doing something like sqlbox also shouldn't be that difficult. On Thu, Mar 24, 2011 at 12:17 PM, Alejandro Guerrieri alejandro.guerri...@gmail.com wrote: I've been playing with kannel and memcached a little bit, but nothing worth posting so far. The unreliable nature of memcached (because of its unreplicated horizontal scaling) doesn't make it very suitable for DLR's in fact (e.g., a node going down and zapping all the dlrs stored in it). Other noSQL engines would make more sense. Regards, Alex On Thu, Mar 24, 2011 at 8:12 PM, Toby Phipps toby.phi...@nexmedia.com.sg wrote: Hi all, Has anyone put any thought into implementing Kannel DLR storage (and the basics of sqlbox MT injection) in a NoSQL datastore? We are using Redis (http://redis.io) for other projects and are very happy with it. Its key-based storage/retrieval and strong support for arrays as atomic queues would make it perfect for Kannel, especially given the transient nature of this data and Redis’ memory speed with disk backing. The main advantages in my mind are that during heavy loads disk access would be minimized, DB polling would disappear (to be replaced with blocking array pops) and the SQL overhead would be removed replaced with set/fetch by keys. No particular reason this would need to be Redis – memcached, MongoDB, Cassandra etc. would also probably work well assuming they have the array/queue support. Any thoughts? It wouldn’t be a small project, but basing the code on the existing implementations of SQL DLRs and sqlbox would mean a big jump start… Toby.
How to start Kannel
I need to some manual for start Kannel because I`m new in Kannel and it`s difficult for me. I hope to understand, thanks.
