Hi,
I went thru old subjects but cannot find any or helpful answers.
We a a unix server with Ingres installed. That server is only used for
php application, or batch using Ingres thru network (Via GCC) (There is
no database on that server).
They are actually 10 GCCs defined with a maxoutbound set to 300 per
GCC. Every day we have a different time "max_Outbound connection
reached" in the errlog.log.

1) Is it a way to get, per GCC, the number of connections used, in real
time?
2) Is it a link between a parameter between the GCN and the GCC?
3) Is it better to have 2 GCCs defined in CBF whith a startup count of
5, or 10 GCCs with a startup count of 1?
4) If only 2 GCCs with 5 at startup, how should they be declared in
"Netutil"?

Thanks for your help.

nionio_usa.

Hi nionio_usa,

> 1) Is it a way to get, per GCC, the number of connections used, in real time?

If you have a full server installation running on the client box, you
might be able to monitor "ima_gcc_info" (in the ingres management
database - "imadb"). The "outbound_current" attribute should tell you
the current connections count for each registered iigcc ).

$ isql imadb
execute procedure ima_set_vnode_domain;
select * from ima_gcc_info;

> 2) Is it a link between a parameter between the GCN and the GCC?

I'm not sure what you mean here. There is a strong link between the
gcn and the gcc, in that it is the gcn that routes the client
connection request to an available gcc if the database name is prefixed
by a vnode. In the case of an outbound connection, the name server will
randomly choose a registered gcc from the name server registry (see
iinamu -- show comsvr). If you have multiple gcc's on the remote
server (which you should, if you are using 10 gcc's x 300 connections
on the client), you will have to have merged connection data entries in
netutil to use them-- one for each remote gcc. (For example, if the
listen address of the remote installation is II, then you should create
n connection entries, with listen addresses II0, II1, II2 ... up to
IIn).

> 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> 5, or 10 GCCs with a startup count of 1?
> 4) If only 2 GCCs with 5 at startup, how should they be declared in
> "Netutil"?

I think it doesn't really matter... as far as I know, the only use for
a duplicate named net server is so you can have an alternate
configuration gcc, which can be started and stopped independently.
Someone please correct me if I'm wrong. I can't think of any reason
that it would benefit you to create duplicate gcc's with a smaller
startup count.

Again, you don't need to do anything specific to use all 10 gcc's on
your client installation... but: are you REALLY getting close to 3000
simultaneous user connections?? Ingres name server load-balancing isn't
very precise, so it's possible that one or another gcc is filling up
before its time, but you still would have to have lots of connections I
think. Does your php application close its connection after doing its
work? The only thing I can imagine that may be happening is that your
getting a lot of hung connections; and that may be happening if you are
overloading the remote server gcc, or if you are experiencing excessive
locking on the server, for example. Check out the server installation
with IPM: do you really have that much dbms traffic? Are connections
hanging around for too long without being released? Verify that you
are using merged netutil entries and that all of the remote gcc's are
being accessed (you should be able to check it with netstat, or use the
imadb query shown above).

If you really do have so much traffic though, another option may be to
bypass the net server altogether, using "connection_type=direct". This
will eliminate considerable overhead of going through the gcc's, and,
as far as I know, has no limit on the number of outbound / inbound
connections.

Open netutil, select the vnode, and choose the "Attributes" option.
Create an attribute "connection_type" with value "direct". Note that
there are restrictions to this parameter: the main one is that the
client and server machines must be the same platform (i.e. Solaris /
Solaris). Also, the Ingres version must be at least II 2.5.

If you really do have ***a lot*** of traffic though
(connects/disconnects), then it's possible your name server may not be
keeping up, and may need some tuning. I think that's an unlikely
possibility though.

Best of luck,

Jim Gramling
Rio de Janeiro

nionio_usa wrote:
> Hi,
> I went thru old subjects but cannot find any or helpful answers.
> We a a unix server with Ingres installed. That server is only used for
> php application, or batch using Ingres thru network (Via GCC) (There is
> no database on that server).
> They are actually 10 GCCs defined with a maxoutbound set to 300 per
> GCC. Every day we have a different time "max_Outbound connection
> reached" in the errlog.log.
>
> 1) Is it a way to get, per GCC, the number of connections used, in real
> time?
> 2) Is it a link between a parameter between the GCN and the GCC?
> 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> 5, or 10 GCCs with a startup count of 1?
> 4) If only 2 GCCs with 5 at startup, how should they be declared in
> "Netutil"?
>
> Thanks for your help.
>
> nionio_usa.

Hi Jim,
Thanks a lot for your help.

I will go thru all of your tests procedure and recommandations.

Thanks.
nionio_usa.


On 24 jan, 02:29, "Jim Gramling" <jimwgraml...@gmail.com> wrote:
> Hi nionio_usa,
>
> > 1) Is it a way to get, per GCC, the number of connections used, in realtime?If you have a full server installation running on the client box, you
> might be able to monitor "ima_gcc_info" (in the ingres management
> database - "imadb"). The "outbound_current" attribute should tell you
> the current connections count for each registered iigcc ).
>
> $ isql imadb
> execute procedure ima_set_vnode_domain;
> select * from ima_gcc_info;
>
> > 2) Is it a link between a parameter between the GCN and the GCC?I'm notsure what you mean here. There is a strong link between the
> gcn and the gcc, in that it is the gcn that routes the client
> connection request to an available gcc if the database name is prefixed
> by a vnode. In the case of an outbound connection, the name server will
> randomly choose a registered gcc from the name server registry (see
> iinamu -- show comsvr). If you have multiple gcc's on the remote
> server (which you should, if you are using 10 gcc's x 300 connections
> on the client), you will have to have merged connection data entries in
> netutil to use them-- one for each remote gcc. (For example, if the
> listen address of the remote installation is II, then you should create
> n connection entries, with listen addresses II0, II1, II2 ... up to
> IIn).
>
> > 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> > 5, or 10 GCCs with a startup count of 1?
> > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > "Netutil"?I think it doesn't really matter... as far as I know, the only use for
> a duplicate named net server is so you can have an alternate
> configuration gcc, which can be started and stopped independently.
> Someone please correct me if I'm wrong. I can't think of any reason
> that it would benefit you to create duplicate gcc's with a smaller
> startup count.
>
> Again, you don't need to do anything specific to use all 10 gcc's on
> your client installation... but: are you REALLY getting close to 3000
> simultaneous user connections?? Ingres name server load-balancing isn't
> very precise, so it's possible that one or another gcc is filling up
> before its time, but you still would have to have lots of connections I
> think. Does your php application close its connection after doing its
> work? The only thing I can imagine that may be happening is that your
> getting a lot of hung connections; and that may be happening if you are
> overloading the remote server gcc, or if you are experiencing excessive
> locking on the server, for example. Check out the server installation
> with IPM: do you really have that much dbms traffic? Are connections
> hanging around for too long without being released? Verify that you
> are using merged netutil entries and that all of the remote gcc's are
> being accessed (you should be able to check it with netstat, or use the
> imadb query shown above).
>
> If you really do have so much traffic though, another option may be to
> bypass the net server altogether, using "connection_type=direct". This
> will eliminate considerable overhead of going through the gcc's, and,
> as far as I know, has no limit on the number of outbound / inbound
> connections.
>
> Open netutil, select the vnode, and choose the "Attributes" option.
> Create an attribute "connection_type" with value "direct". Note that
> there are restrictions to this parameter: the main one is that the
> client and server machines must be the same platform (i.e. Solaris /
> Solaris). Also, the Ingres version must be at least II 2.5.
>
> If you really do have ***a lot*** of traffic though
> (connects/disconnects), then it's possible your name server may not be
> keeping up, and may need some tuning. I think that's an unlikely
> possibility though.
>
> Best of luck,
>
> Jim Gramling
> Rio de Janeiro
>
>
>
> nionio_usa wrote:
> > Hi,
> > I went thru old subjects but cannot find any or helpful answers.
> > We a a unix server with Ingres installed. That server is only used for
> > php application, or batch using Ingres thru network (Via GCC) (There is
> > no database on that server).
> > They are actually 10 GCCs defined with a maxoutbound set to 300 per
> > GCC. Every day we have a different time "max_Outbound connection
> > reached" in the errlog.log.
>
> > 1) Is it a way to get, per GCC, the number of connections used, in real
> > time?
> > 2) Is it a link between a parameter between the GCN and the GCC?
> > 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> > 5, or 10 GCCs with a startup count of 1?
> > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > "Netutil"?
>
> > Thanks for your help.
>
> > nionio_usa.- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -

Hi again,
I ran the select on the ima_gcc_info but I surprise to see that the
current and the max values are divided by 10 compare to what's defined
in CBF?

|max |cur |net_server |
|0000000030|0000000013|clust7::/@64629 |
|0000000030|0000000014|clust7::/@64634 |
|0000000030|0000000013|clust7::/@64649 |
|0000000030|0000000013|clust7::/@64665 |
|0000000030|0000000014|clust7::/@64668 |
|0000000030|0000000013|clust7::/@64673 |
|0000000030|0000000014|clust7::/@64676 |
|0000000030|0000000013|clust7::/@64678 |
|0000000050|0000000014|clust7::/@64682 |

In CBF, value for outbound limit is either 300 or 500.

Any Idea?

Thanks.

nionio_usa
On 24 jan, 10:07, "nionio_usa" <antonio.bote...@arc-intl.com> wrote:
> Hi Jim,
> Thanks a lot for your help.
>
> I will go thru all of your tests procedure and recommandations.
>
> Thanks.
> nionio_usa.
>
> On 24 jan, 02:29, "Jim Gramling" <jimwgraml...@gmail.com> wrote:
>
>
>
> > Hi nionio_usa,
>
> > > 1) Is it a way to get, per GCC, the number of connections used, in real time?If you have a full server installation running on the client box, you
> > might be able to monitor "ima_gcc_info" (in the ingres management
> > database - "imadb"). The "outbound_current" attribute should tell you
> > the current connections count for each registered iigcc ).
>
> > $ isql imadb
> > execute procedure ima_set_vnode_domain;
> > select * fromima_gcc_info;
>
> > > 2) Is it a link between a parameter between the GCN and the GCC?I'm not sure what you mean here. There is a strong link between the
> > gcn and the gcc, in that it is the gcn that routes the client
> > connection request to an available gcc if the database name is prefixed
> > by a vnode. In the case of an outbound connection, the name server will
> > randomly choose a registered gcc from the name server registry (see
> > iinamu -- show comsvr). If you have multiple gcc's on the remote
> > server (which you should, if you are using 10 gcc's x 300 connections
> > on the client), you will have to have merged connection data entries in
> > netutil to use them-- one for each remote gcc. (For example, if the
> > listen address of the remote installation is II, then you should create
> > n connection entries, with listen addresses II0, II1, II2 ... up to
> > IIn).
>
> > > 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> > > 5, or 10 GCCs with a startup count of 1?
> > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > "Netutil"?I think it doesn't really matter... as far as I know, the only use for
> > a duplicate named net server is so you can have an alternate
> > configuration gcc, which can be started and stopped independently.
> > Someone please correct me if I'm wrong. I can't think of any reason
> > that it would benefit you to create duplicate gcc's with a smaller
> > startup count.
>
> > Again, you don't need to do anything specific to use all 10 gcc's on
> > your client installation... but: are you REALLY getting close to 3000
> > simultaneous user connections?? Ingres name server load-balancing isn't
> > very precise, so it's possible that one or another gcc is filling up
> > before its time, but you still would have to have lots of connections I
> > think. Does your php application close its connection after doing its
> > work? The only thing I can imagine that may be happening is that your
> > getting a lot of hung connections; and that may be happening if you are
> > overloading the remote server gcc, or if you are experiencing excessive
> > locking on the server, for example. Check out the server installation
> > with IPM: do you really have that much dbms traffic? Are connections
> > hanging around for too long without being released? Verify that you
> > are using merged netutil entries and that all of the remote gcc's are
> > being accessed (you should be able to check it with netstat, or use the
> > imadb query shown above).
>
> > If you really do have so much traffic though, another option may be to
> > bypass the net server altogether, using "connection_type=direct". This
> > will eliminate considerable overhead of going through the gcc's, and,
> > as far as I know, has no limit on the number of outbound / inbound
> > connections.
>
> > Open netutil, select the vnode, and choose the "Attributes" option.
> > Create an attribute "connection_type" with value "direct". Note that
> > there are restrictions to this parameter: the main one is that the
> > client and server machines must be the same platform (i.e. Solaris /
> > Solaris). Also, the Ingres version must be at least II 2.5.
>
> > If you really do have ***a lot*** of traffic though
> > (connects/disconnects), then it's possible your name server may not be
> > keeping up, and may need some tuning. I think that's an unlikely
> > possibility though.
>
> > Best of luck,
>
> > Jim Gramling
> > Rio de Janeiro
>
> > nionio_usa wrote:
> > > Hi,
> > > I went thru old subjects but cannot find any or helpful answers.
> > > We a a unix server with Ingres installed. That server is only used for
> > > php application, or batch using Ingres thru network (Via GCC) (There is
> > > no database on that server).
> > > They are actually 10 GCCs defined with a maxoutbound set to 300 per
> > > GCC. Every day we have a different time "max_Outbound connection
> > > reached" in the errlog.log.
>
> > > 1) Is it a way to get, per GCC, the number of connections used, in real
> > > time?
> > > 2) Is it a link between a parameter between the GCN and the GCC?
> > > 3) Is it better to have 2 GCCs defined in CBF whith a startup count of
> > > 5, or 10 GCCs with a startup count of 1?
> > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > "Netutil"?
>
> > > Thanks for your help.
>
> > > nionio_usa.- Masquer le texte des messages précédents -- Afficherle texte des messages précédents -- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -

Yeah, I've seen that before Ninio. For some reason, the last
character is being truncated, which means you can only get a general
idea of how many connections there are in each gcc ... one-hundred-and-
thirty something, one-hundred-and-forty-something ...
It's strange, but probably enough to tell you what you need to know.

Did you check out the same query on the server side? (look at
inbound ...)

Regards,
Jim Gramling
Rio de Janeiro

On Jan 30, 9:53 am, "nionio_usa" <antonio.bote...@arc-intl.com> wrote:
> Hi again,
> I ran the select on the ima_gcc_info but I surprise to see that the
> current and the max values are divided by 10 compare to what's defined
> in CBF?
>
> |max |cur |net_server |
> |0000000030|0000000013|clust7::/@64629 |
> |0000000030|0000000014|clust7::/@64634 |
> |0000000030|0000000013|clust7::/@64649 |
> |0000000030|0000000013|clust7::/@64665 |
> |0000000030|0000000014|clust7::/@64668 |
> |0000000030|0000000013|clust7::/@64673 |
> |0000000030|0000000014|clust7::/@64676 |
> |0000000030|0000000013|clust7::/@64678 |
> |0000000050|0000000014|clust7::/@64682 |
>
> In CBF, value for outbound limit is either 300 or 500.
>
> Any Idea?
>
> Thanks.
>
> nionio_usa
> On 24 jan, 10:07, "nionio_usa" <antonio.bote...@arc-intl.com> wrote:
>
>
>
> > Hi Jim,
> > Thanks a lot for your help.
>
> > I will go thru all of your tests procedure and recommandations.
>
> > Thanks.
> > nionio_usa.
>
> > On 24 jan, 02:29, "Jim Gramling" <jimwgraml...@gmail.com> wrote:
>
> > > Hi nionio_usa,
>
> > > > 1) Is it a way to get, per GCC, the number of connections used, in real time?If you have a full server installation running on the client box,you
> > > might be able to monitor "ima_gcc_info" (in the ingres management
> > > database - "imadb"). The "outbound_current" attribute should tell you
> > > the current connections count for each registered iigcc ).
>
> > > $ isql imadb
> > > execute procedure ima_set_vnode_domain;
> > > select * fromima_gcc_info;
>
> > > > 2) Is it a link between a parameter between the GCN and the GCC?I'mnot sure what you mean here. There is a strong link between the
> > > gcn and the gcc, in that it is the gcn that routes the client
> > > connection request to an available gcc if the database name is prefixed
> > > by a vnode. In the case of an outbound connection, the name server will
> > > randomly choose a registered gcc from the name server registry (see
> > > iinamu -- show comsvr). If you have multiple gcc's on the remote
> > > server (which you should, if you are using 10 gcc's x 300 connections
> > > on the client), you will have to have merged connection data entries in
> > > netutil to use them-- one for each remote gcc. (For example, if the
> > > listen address of the remote installation is II, then you should create
> > > n connection entries, with listen addresses II0, II1, II2 ... up to
> > > IIn).
>
> > > > 3) Is it better to have 2 GCCs defined in CBF whith a startup countof
> > > > 5, or 10 GCCs with a startup count of 1?
> > > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > > "Netutil"?I think it doesn't really matter... as far as I know, the only use for
> > > a duplicate named net server is so you can have an alternate
> > > configuration gcc, which can be started and stopped independently.
> > > Someone please correct me if I'm wrong. I can't think of any reason
> > > that it would benefit you to create duplicate gcc's with a smaller
> > > startup count.
>
> > > Again, you don't need to do anything specific to use all 10 gcc's on
> > > your client installation... but: are you REALLY getting close to 3000
> > > simultaneous user connections?? Ingres name server load-balancing isn't
> > > very precise, so it's possible that one or another gcc is filling up
> > > before its time, but you still would have to have lots of connectionsI
> > > think. Does your php application close its connection after doing its
> > > work? The only thing I can imagine that may be happening is that your
> > > getting a lot of hung connections; and that may be happening if you are
> > > overloading the remote server gcc, or if you are experiencing excessive
> > > locking on the server, for example. Check out the server installation
> > > with IPM: do you really have that much dbms traffic? Are connections
> > > hanging around for too long without being released? Verify that you
> > > are using merged netutil entries and that all of the remote gcc's are
> > > being accessed (you should be able to check it with netstat, or use the
> > > imadb query shown above).
>
> > > If you really do have so much traffic though, another option may be to
> > > bypass the net server altogether, using "connection_type=direct". This
> > > will eliminate considerable overhead of going through the gcc's, and,
> > > as far as I know, has no limit on the number of outbound / inbound
> > > connections.
>
> > > Open netutil, select the vnode, and choose the "Attributes" option.
> > > Create an attribute "connection_type" with value "direct". Note that
> > > there are restrictions to this parameter: the main one is that the
> > > client and server machines must be the same platform (i.e. Solaris /
> > > Solaris). Also, the Ingres version must be at least II 2.5.
>
> > > If you really do have ***a lot*** of traffic though
> > > (connects/disconnects), then it's possible your name server may not be
> > > keeping up, and may need some tuning. I think that's an unlikely
> > > possibility though.
>
> > > Best of luck,
>
> > > Jim Gramling
> > > Rio de Janeiro
>
> > > nionio_usa wrote:
> > > > Hi,
> > > > I went thru old subjects but cannot find any or helpful answers.
> > > > We a a unix server with Ingres installed. That server is only used for
> > > > php application, or batch using Ingres thru network (Via GCC) (There is
> > > > no database on that server).
> > > > They are actually 10 GCCs defined with a maxoutbound set to 300 per
> > > > GCC. Every day we have a different time "max_Outbound connection
> > > > reached" in the errlog.log.
>
> > > > 1) Is it a way to get, per GCC, the number of connections used, in real
> > > > time?
> > > > 2) Is it a link between a parameter between the GCN and the GCC?
> > > > 3) Is it better to have 2 GCCs defined in CBF whith a startup countof
> > > > 5, or 10 GCCs with a startup count of 1?
> > > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > > "Netutil"?
>
> > > > Thanks for your help.
>
> > > > nionio_usa.- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -- Hide quoted text -- Show quoted text -

Ninio, another thing I just noticed: I'm not sure that there is much
point in having gcc's with 500 outbound connections and others with
300 outbound. I may be wrong, but I'm under the impression that the
gcn isn't smart enough to know that a particular gcc still has
available outbound slots, which means that your smaller configurations
will start running out of sessions even though the higher-config ones
still have plenty of slots.

Regards,
Jim Gramling
Rio de Janeiro

On Jan 30, 9:53 am, "nionio_usa" <antonio.bote...@arc-intl.com> wrote:
> Hi again,
> I ran the select on the ima_gcc_info but I surprise to see that the
> current and the max values are divided by 10 compare to what's defined
> in CBF?
>
> |max |cur |net_server |
> |0000000030|0000000013|clust7::/@64629 |
> |0000000030|0000000014|clust7::/@64634 |
> |0000000030|0000000013|clust7::/@64649 |
> |0000000030|0000000013|clust7::/@64665 |
> |0000000030|0000000014|clust7::/@64668 |
> |0000000030|0000000013|clust7::/@64673 |
> |0000000030|0000000014|clust7::/@64676 |
> |0000000030|0000000013|clust7::/@64678 |
> |0000000050|0000000014|clust7::/@64682 |
>
> In CBF, value for outbound limit is either 300 or 500.
>
> Any Idea?
>
> Thanks.
>
> nionio_usa
> On 24 jan, 10:07, "nionio_usa" <antonio.bote...@arc-intl.com> wrote:
>
>
>
> > Hi Jim,
> > Thanks a lot for your help.
>
> > I will go thru all of your tests procedure and recommandations.
>
> > Thanks.
> > nionio_usa.
>
> > On 24 jan, 02:29, "Jim Gramling" <jimwgraml...@gmail.com> wrote:
>
> > > Hi nionio_usa,
>
> > > > 1) Is it a way to get, per GCC, the number of connections used, in real time?If you have a full server installation running on the client box,you
> > > might be able to monitor "ima_gcc_info" (in the ingres management
> > > database - "imadb"). The "outbound_current" attribute should tell you
> > > the current connections count for each registered iigcc ).
>
> > > $ isql imadb
> > > execute procedure ima_set_vnode_domain;
> > > select * fromima_gcc_info;
>
> > > > 2) Is it a link between a parameter between the GCN and the GCC?I'mnot sure what you mean here. There is a strong link between the
> > > gcn and the gcc, in that it is the gcn that routes the client
> > > connection request to an available gcc if the database name is prefixed
> > > by a vnode. In the case of an outbound connection, the name server will
> > > randomly choose a registered gcc from the name server registry (see
> > > iinamu -- show comsvr). If you have multiple gcc's on the remote
> > > server (which you should, if you are using 10 gcc's x 300 connections
> > > on the client), you will have to have merged connection data entries in
> > > netutil to use them-- one for each remote gcc. (For example, if the
> > > listen address of the remote installation is II, then you should create
> > > n connection entries, with listen addresses II0, II1, II2 ... up to
> > > IIn).
>
> > > > 3) Is it better to have 2 GCCs defined in CBF whith a startup countof
> > > > 5, or 10 GCCs with a startup count of 1?
> > > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > > "Netutil"?I think it doesn't really matter... as far as I know, the only use for
> > > a duplicate named net server is so you can have an alternate
> > > configuration gcc, which can be started and stopped independently.
> > > Someone please correct me if I'm wrong. I can't think of any reason
> > > that it would benefit you to create duplicate gcc's with a smaller
> > > startup count.
>
> > > Again, you don't need to do anything specific to use all 10 gcc's on
> > > your client installation... but: are you REALLY getting close to 3000
> > > simultaneous user connections?? Ingres name server load-balancing isn't
> > > very precise, so it's possible that one or another gcc is filling up
> > > before its time, but you still would have to have lots of connectionsI
> > > think. Does your php application close its connection after doing its
> > > work? The only thing I can imagine that may be happening is that your
> > > getting a lot of hung connections; and that may be happening if you are
> > > overloading the remote server gcc, or if you are experiencing excessive
> > > locking on the server, for example. Check out the server installation
> > > with IPM: do you really have that much dbms traffic? Are connections
> > > hanging around for too long without being released? Verify that you
> > > are using merged netutil entries and that all of the remote gcc's are
> > > being accessed (you should be able to check it with netstat, or use the
> > > imadb query shown above).
>
> > > If you really do have so much traffic though, another option may be to
> > > bypass the net server altogether, using "connection_type=direct". This
> > > will eliminate considerable overhead of going through the gcc's, and,
> > > as far as I know, has no limit on the number of outbound / inbound
> > > connections.
>
> > > Open netutil, select the vnode, and choose the "Attributes" option.
> > > Create an attribute "connection_type" with value "direct". Note that
> > > there are restrictions to this parameter: the main one is that the
> > > client and server machines must be the same platform (i.e. Solaris /
> > > Solaris). Also, the Ingres version must be at least II 2.5.
>
> > > If you really do have ***a lot*** of traffic though
> > > (connects/disconnects), then it's possible your name server may not be
> > > keeping up, and may need some tuning. I think that's an unlikely
> > > possibility though.
>
> > > Best of luck,
>
> > > Jim Gramling
> > > Rio de Janeiro
>
> > > nionio_usa wrote:
> > > > Hi,
> > > > I went thru old subjects but cannot find any or helpful answers.
> > > > We a a unix server with Ingres installed. That server is only used for
> > > > php application, or batch using Ingres thru network (Via GCC) (There is
> > > > no database on that server).
> > > > They are actually 10 GCCs defined with a maxoutbound set to 300 per
> > > > GCC. Every day we have a different time "max_Outbound connection
> > > > reached" in the errlog.log.
>
> > > > 1) Is it a way to get, per GCC, the number of connections used, in real
> > > > time?
> > > > 2) Is it a link between a parameter between the GCN and the GCC?
> > > > 3) Is it better to have 2 GCCs defined in CBF whith a startup countof
> > > > 5, or 10 GCCs with a startup count of 1?
> > > > 4) If only 2 GCCs with 5 at startup, how should they be declared in
> > > > "Netutil"?
>
> > > > Thanks for your help.
>
> > > > nionio_usa.- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -- Masquer le texte des messages précédents -- Afficher le texte des messages précédents -- Hide quoted text -- Show quoted text -