any idea what could have changed from last week until this week. Did the network change? did the size of the packets change? If the clients disconnect while syncing then this will cause overhead in the system. Have you logged at the log files on the client side?
first of all many thanks for your fast answer .
Unfortunatelly the most logged error code on client side ist that:
CNS-9107 Network disconnected during sync.
Action in the Mobile Server Documentation is: Check your Mobile Server URL; make sure that you have an active network connection.
We weren't sure if this error was thrown because of the slow performance of the Mobile Server, so many clients got declined, or if it is the other way round.
Could it also be that a slow network is the problem and thats the reason for the problems on the mobile Server?
Our customer simly says, everything in their network is OK and they changed nothing. We dont have the ability to check that by ourselfs.
the client is receiving a network error. This is something we cannot control in the client code as it is occurring external to it. What may have happened is the load on the customer's network has increased which is now causing it to overload. You can look at this more deeply by using a tool that can analyze the network traffic. We have seen in our own internal testing that when you overuse the network bandwidth it can cause performance to degrade. Maybe you can get the customer to do some type of network timing test. The client cannot sync if the network is unstable. In this case it is best to wait until the network is more stable. With a client side db, it can store a lot of information before it has to sync. Another option is change the type of network used for sync -- if currently using wireless maybe try wired.
many thanks for your proposals. The MGP runs every minute like the default settings are, because our application is a time sensitive. The data have to be very fast on the clients in the nights and thex are only a few minutes before available. I think the parameters are important in environemnts where the mgps runs only after a few minutes?
Let me summarize the situation since my las post:
We have days (nights) where everything goes quite well. Mainly that is thrusday and friday. But on the other hand, there are days where everything really is a mess.
The mgp caycles increase and when you look to mobile manager you see 70 clients syncing very very slowly (about 300 - 600 seconds) for just a little bit of data. Then we get the disonnects and the messages in the err.log.
What do you think, is this somehow network related, or ist this a logical problem in the Mobile Server itself?
Like postes earlier, there are 500 clients syncing against a standalone mobile server.
We just couldnt figure out, why the mgp times somehow and sometimes increase that much, that there are these problems.
First we thought, that they sync newly over wireless (GPRS or Edge) and that syncing is that slowly, that everything goes down, but arent sure. Is this possible because of the architecture of the Mobile Server? That a very slow network can slow down the mgps and that does timeouts and all the other stuff?
How many clients can a Standalone Mobile Server normally serve at same time without performance problems. What do you thin about the hardware I wrote? Is it powerful enough?
what is different on Thursday and Friday? Are the number of clients different on the other days? Is the load different, etc.
This all appears to be network related.
In regard to the machine you talked about originally, are you talking about the size of the machine where the backend database is running? Are you talking about the machine where the application server is running? Have you looked at any kind of performance stats on that machine. It is very difficult to say if you machine is powerful enough. You have to gather performance data on the machine while it is running and have a look. Let me give you an example: A car take 30 seconds to go 1 mile on flat land, so how long does it take if the mile is up a hill with a 40% incline, how long does it take if it is a 60% incline? The flat vs 40% vs 60% incline represents network load. You need to analyze your environment. In the example, maybe the car is not strong enough to go up a 60% incline and it takes hours.
A possible root cause, it is that your customer has some mobile client which didn't synchronized for a while, as the repository keeps information for them, it can decrease the MGP performance.
Check if you are in this case, and synchronize or delete the inactive mobile client.