Hi, There are a number of ways to tackle this problem, however I have solved this by using VMware and VMware Fault Tolerance. VMware Fault FT is a technology based on vLockStep that provides continuous availability in the event of physical ESX server failure by creating a shadow virtual machine which is in step with the primary machine. What this means for Squid is that you are running a single instance of Squid on your ESX(i) cluster in FT protected mode. Should the physical host fail, the Squid instance shadow copy kicks in with near instantaneous effect. No connections / interruptions to traffic occurs, and the end users are none the wiser as all connections are persistent through the outage. The limitation of VMware FT however is that it constrained to one vCPU - but for Squid I don't think that this is a massive issue. Solving the problem of lockstep for 2 single CPU virtual machines is much simpler than multi CPU virtual machines. The underlying technology for vLockstep uses record / replay to record the indeterministic input from one VM and replay it on the second VM over a specific configured logging network. While this works great for single CPU's in vSMP environments shared memory access by the CPU's is not deterministic - hence recording only inputs from one virtual machine and replaying them on another machine does not guarantee deterministic replay to exactly replicate the original environment. This makes FT for vSMP distinctly harder and a much more different problem to solve. It is unknown at this point in time as to when vSMP support will be available for VMware FT - even with the release of VMware ESX(i) 4.1 there is no change in this arena. If you are using VMware within your environment, have a look at VMware FT. I do have video's demonstrating FT in operation when dealing with web content highlighting the persistence of connections during failover. Best Regards, Alex -----Original Message----- From: viswa [mailto:waytoviswa@xxxxxxxxx] Sent: Wednesday, 21 July 2010 4:48 PM To: squid-users@xxxxxxxxxxxxxxx Subject: squid mirror hi all i am using two squid servers with HA Cluster. Always one squid server is responds to clients another one for backup. If the squid one is down the another squid server responds to clients.And my problem is persistent connection . is this possible to continue to client persistent connection after the squid one failure ?? is possible to mirror squid connections to another squid?? Thanks
