Replication traffic on the network has spiked up !!!
Posted in 2003
User on IDS 9.21.UC5 (Solaris 8) saw a spike in traffic on the replication port (1530) after tuning LRU_MAX_DIRTY, LRU_MIN_DIRTY, NUMAIOVPS, CLEANERS and LRUAGE, and asked whether those changes could be responsible. Madison Pruet helped confirm it was HDR (onstat -g dri; onstat -g ddr indicates ER) and explained HDR sends only log buffer pages, ACKs and occasional pings, so more traffic implies more logging — check the message log for "log complete" entries — and noted client read traffic to the secondary can be mistaken for replication unless separate listener ports are used. He saw no way the parameter changes could cause it. No definitive root cause is recorded.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Topics: General Discussion
HDR or ER?
How are you able to monitor replication traffic?
M.P.
"Jack A" <hrehal@yahoo.com> wrote in message
news:5eef134e5c079abca0ea83d4dc748474@free.teranews.com...
> Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS 5.8. Last
> week we changed the following configuration parameters
>
> LRU_MAX_DIRTY - 30 to 5
> LRU_MIN_DIRTY - 20 to 2
> NUMAIOVPS - 2 to 6
> CLEANERS - 8 to 16
> LRUAGE - 1>
> Common sense suggests that network traffic in replication would depend
only
> on the amount of data modifed on the primary. Is it possible that any of
> there parameter changes can cause too much replication traffic ?
>
> Thanks in advance
>
>
Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS 5.8. Last
week we changed the following configuration parameters
LRU_MAX_DIRTY - 30 to 5
LRU_MIN_DIRTY - 20 to 2
NUMAIOVPS - 2 to 6
CLEANERS - 8 to 16
LRUAGE - 1
Common sense suggests that network traffic in replication would depend only
on the amount of data modifed on the primary. Is it possible that any of
there parameter changes can cause too much replication traffic ?
Thanks in advance
I'll answer this question with another question. I think it is HDR. Any way
of being 100% sure. How do I check ?
The network guys isolated the port to which the traffic was found to be
going. Its TCP Port 1530 which is supposed to be the informix replication
port.
"Madison Pruet" <mpruet@comcast.net> wrote in message
news:WIcYa.55982$It4.31506@rwcrnsc51.ops.asp.att.net...
> HDR or ER?
>
> How are you able to monitor replication traffic?
>
> M.P.
>
> "Jack A" <hrehal@yahoo.com> wrote in message
> news:5eef134e5c079abca0ea83d4dc748474@free.teranews.com...
> > Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS 5.8.
Last
> > week we changed the following configuration parameters
> >
> > LRU_MAX_DIRTY - 30 to 5
> > LRU_MIN_DIRTY - 20 to 2
> > NUMAIOVPS - 2 to 6
> > CLEANERS - 8 to 16
> > LRUAGE - 1> >
> > Common sense suggests that network traffic in replication would depend
> only
> > on the amount of data modifed on the primary. Is it possible that any of
> > there parameter changes can cause too much replication traffic ?
> >
> > Thanks in advance
> >
> >
>
>
run onstat -g ddr
If it produces output, then you are running with ER(CDR)
onstat -g dri will show the HDR state.
It's a little difficult to isolate HDR traffic from normal client/server
traffic. That's why I asked the question.
M.Pruet
"Jack A" <hrehal@yahoo.com> wrote in message
news:2ae7a70cc141d60ffa90046e01603107@free.teranews.com...
> I'll answer this question with another question. I think it is HDR. Any
way
> of being 100% sure. How do I check ?
>
> The network guys isolated the port to which the traffic was found to be
> going. Its TCP Port 1530 which is supposed to be the informix replication
> port.
>
> "Madison Pruet" <mpruet@comcast.net> wrote in message
> news:WIcYa.55982$It4.31506@rwcrnsc51.ops.asp.att.net...
> > HDR or ER?
> >
> > How are you able to monitor replication traffic?
> >
> > M.P.
> >
> > "Jack A" <hrehal@yahoo.com> wrote in message
> > news:5eef134e5c079abca0ea83d4dc748474@free.teranews.com...
> > > Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS 5.8.
> Last
> > > week we changed the following configuration parameters
> > >
> > > LRU_MAX_DIRTY - 30 to 5
> > > LRU_MIN_DIRTY - 20 to 2
> > > NUMAIOVPS - 2 to 6
> > > CLEANERS - 8 to 16
> > > LRUAGE - 1> > >
> > > Common sense suggests that network traffic in replication would depend
> > only
> > > on the amount of data modifed on the primary. Is it possible that any
of
> > > there parameter changes can cause too much replication traffic ?
> > >
> > > Thanks in advance
> > >
> > >
> >
> >
>
>
It is HDR replication. Heres the output
onstat -g dri
Informix Dynamic Server 2000 Version 9.21.UC5 -- On-Line (Prim) -- Up 6
days
20:14:28 -- 112640 Kbytes
Data Replication:
Type State Paired server Last DR CKPT (id/pg)
primary on triptik_slv 3708 / 484
DRINTERVAL 30
DRTIMEOUT 30
DRLOSTFOUND /opt/ids2k/v9r21uc5/etc/dr.lostfound
"Madison Pruet" <mpruet@comcast.net> wrote in message
news:FEeYa.56591$It4.31844@rwcrnsc51.ops.asp.att.net...
> run onstat -g ddr
> If it produces output, then you are running with ER(CDR)
>
> onstat -g dri will show the HDR state.>
> It's a little difficult to isolate HDR traffic from normal client/server
> traffic. That's why I asked the question.
>
> M.Pruet
>
> "Jack A" <hrehal@yahoo.com> wrote in message
> news:2ae7a70cc141d60ffa90046e01603107@free.teranews.com...
> > I'll answer this question with another question. I think it is HDR. Any
> way
> > of being 100% sure. How do I check ?
> >
> > The network guys isolated the port to which the traffic was found to be
> > going. Its TCP Port 1530 which is supposed to be the informix
replication
> > port.
> >
> > "Madison Pruet" <mpruet@comcast.net> wrote in message
> > news:WIcYa.55982$It4.31506@rwcrnsc51.ops.asp.att.net...
> > > HDR or ER?
> > >
> > > How are you able to monitor replication traffic?
> > >
> > > M.P.
> > >
> > > "Jack A" <hrehal@yahoo.com> wrote in message
> > > news:5eef134e5c079abca0ea83d4dc748474@free.teranews.com...
> > > > Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS 5.8.
> > Last
> > > > week we changed the following configuration parameters
> > > >
> > > > LRU_MAX_DIRTY - 30 to 5
> > > > LRU_MIN_DIRTY - 20 to 2
> > > > NUMAIOVPS - 2 to 6
> > > > CLEANERS - 8 to 16
> > > > LRUAGE - 1> > > >
> > > > Common sense suggests that network traffic in replication would
depend
> > > only
> > > > on the amount of data modifed on the primary. Is it possible that
any
> of
> > > > there parameter changes can cause too much replication traffic ?
> > > >
> > > > Thanks in advance
> > > >
> > > >
> > >
> > >
> >
> >
>
>
Jack,
With HDR the only thing that is transmitted between the source/target are
the log buffer pages, ACKs that the log buffers have been received, and an
occasional PING message. If there is an increase in network traffic between
the primary/secondary and it is on a port that is dedicated to the
replication traffic, then there should be a corresponding increase in log
usage. You can get an idea on the log file usage by examining the message
log file for the "log complete" messages.
Be aware that the HDR secondary server can be used by clients for read
purposes. Unless you have two ports (listener threads) on the secondary
server, it will not be possible to distinguish the replication traffic from
client traffic that is running reports on the secondary server.
I can think of no reason that changes that you described would cause an
increase in replication network traffic.
"Jack A" <hrehal@yahoo.com> wrote in message
news:e2b26ce5160a9f9adf6631023e00c406@free.teranews.com...
> It is HDR replication. Heres the output
>
> onstat -g dri>
> Informix Dynamic Server 2000 Version 9.21.UC5 -- On-Line (Prim) -- Up
6
> days
> 20:14:28 -- 112640 Kbytes
>
> Data Replication:
> Type State Paired server Last DR CKPT (id/pg)
> primary on triptik_slv 3708 / 484
>
> DRINTERVAL 30
> DRTIMEOUT 30
> DRLOSTFOUND /opt/ids2k/v9r21uc5/etc/dr.lostfound>
>
> "Madison Pruet" <mpruet@comcast.net> wrote in message
> news:FEeYa.56591$It4.31844@rwcrnsc51.ops.asp.att.net...
> > run onstat -g ddr
> > If it produces output, then you are running with ER(CDR)
> >
> > onstat -g dri will show the HDR state.> >
> > It's a little difficult to isolate HDR traffic from normal client/server
> > traffic. That's why I asked the question.
> >
> > M.Pruet
> >
> > "Jack A" <hrehal@yahoo.com> wrote in message
> > news:2ae7a70cc141d60ffa90046e01603107@free.teranews.com...
> > > I'll answer this question with another question. I think it is HDR.
Any
> > way
> > > of being 100% sure. How do I check ?
> > >
> > > The network guys isolated the port to which the traffic was found to
be
> > > going. Its TCP Port 1530 which is supposed to be the informix
> replication
> > > port.
> > >
> > > "Madison Pruet" <mpruet@comcast.net> wrote in message
> > > news:WIcYa.55982$It4.31506@rwcrnsc51.ops.asp.att.net...
> > > > HDR or ER?
> > > >
> > > > How are you able to monitor replication traffic?
> > > >
> > > > M.P.
> > > >
> > > > "Jack A" <hrehal@yahoo.com> wrote in message
> > > > news:5eef134e5c079abca0ea83d4dc748474@free.teranews.com...
> > > > > Running Informix Dynamic Server 2000 Version 9.21.UC5 on SunOS
5.8.
> > > Last
> > > > > week we changed the following configuration parameters
> > > > >
> > > > > LRU_MAX_DIRTY - 30 to 5
> > > > > LRU_MIN_DIRTY - 20 to 2
> > > > > NUMAIOVPS - 2 to 6
> > > > > CLEANERS - 8 to 16
> > > > > LRUAGE - 1> > > > >
> > > > > Common sense suggests that network traffic in replication would
> depend
> > > > only
> > > > > on the amount of data modifed on the primary. Is it possible that
> any
> > of
> > > > > there parameter changes can cause too much replication traffic ?
> > > > >
> > > > > Thanks in advance
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>