Re: ER to replace HDR in HA environment
Paul, Any idea who has set your clients expectations that ER is the answer to all their woes? I have a couple of cncerns / comments with what the expect to achieve moving to ER! I am concerned that it will take a load of work, without addressing all the issues the client has. > > -- for planned maintenance, we typically need two 5-10 minute outages to > > switch secondary to standard mode and move the client connections over -- > > client wants these outages, plus all unplanned outages, eliminated > > completely - figures that ER is the answer The client think ER will elimiate all planned outages? Its easy to eliminate planned outages. Don't change your database schema in any way. Don't upgrade your database in any way. Don't update your OS. But seriously, I'm not sure what the client will see here that they don't see at the moment with HDR. Maybe Madisson will shoot me down in flames, but ER won't redirect network connections from one server to other other when the server goes down any more than HDR will. You will still need to stop communication with their server and redirect the connections to the other server. So the impact will be reduced, but not eliminated! You may want to consider IP aliases as a better method to achieve seamless connections to the database if you are not doing so already. Now, their other requirement, eliminated unplanned outages completely?? estebanc00@yahoo.com.ar (Esteban Casuscelli) wrote in message news:<8a8f14b9.0402031647.5e035d09@posting.google.com>... > How many tables are you going to replicate? > Schema changes will affect you production enviroment, ER is based on > the schema for each table. > Also you will have a lot of administration for an update anywhere env. > rather than HDR. > a big number of logical logs if it is a high txn volume. > > esteban.- > > mosserp@wellsfargo.com wrote in message news:<bvoo8m$guc$1@terabinaries.xmission.com>... > > We have been told to figure out how to replace our current HDR environment > > with an ER update-anywhere solution, in order to eliminate "outages" > > (planned or unplanned) with the DB server. > > > > Our environment: > > -- HP-UX 11.0 on identical HP RP7400 servers (8 cpu at 750 MHz, 16Gb memory) > > -- IDS 9.30.FC4XC > > -- servers are located hundreds of miles from each other > > -- high txn volume > > -- HA is *extremely* critical, no regular maintenance window > > -- schema changes occur 2-4 times per year > > -- for planned maintenance, we typically need two 5-10 minute outages to > > switch secondary to standard mode and move the client connections over -- > > client wants these outages, plus all unplanned outages, eliminated > > completely - figures that ER is the answer > > > > Has anyone else implemented ER in update-anywhere (aka, peer-to-peer) mode > > in a high-volume, extremely high-availability environment? If so, what > > should we watch out for, etc.? > > > > Comments? Recommendations? > > > > TIA, > > Paul Mosser > > sending to informix-list > > sending to informix-list
Nope - not going to shoot you down in flames. I've already communicated with Paul about this. While ER can be used for availability solutions, it is not as simple to use as HDR. ER, however, works better then HDR if you want to 1) use a radically different system for the backup, 2) need only a partial set of your data for recoverability, or 3) want to use the secondary for more than dirty read reporting. I view ER more of a distributed solution, rather than an availability solution. That's why we implemented ER/HDR in 9.4. M.P. "scottishpoet" <dryburghj@yahoo.com> wrote in message news:81714288.0402040732.43d51a85@posting.google.com... > Paul, > > Any idea who has set your clients expectations that ER is the answer > to all their woes? > > I have a couple of cncerns / comments with what the expect to achieve > moving to ER! I am concerned that it will take a load of work, without > addressing all the issues the client has. > > > > -- for planned maintenance, we typically need two 5-10 minute outages to > > > switch secondary to standard mode and move the client connections over -- > > > client wants these outages, plus all unplanned outages, eliminated > > > completely - figures that ER is the answer > > The client think ER will elimiate all planned outages? > > Its easy to eliminate planned outages. Don't change your database > schema in any way. Don't upgrade your database in any way. Don't > update your OS. > > But seriously, I'm not sure what the client will see here that they > don't see at the moment with HDR. Maybe Madisson will shoot me down in > flames, but ER won't redirect network connections from one server to > other other when the server goes down any more than HDR will. You will > still need to stop communication with their server and redirect the > connections to the other server. So the impact will be reduced, but > not eliminated! > > You may want to consider IP aliases as a better method to achieve > seamless connections to the database if you are not doing so already. > > Now, their other requirement, eliminated unplanned outages > completely?? > > > > > > estebanc00@yahoo.com.ar (Esteban Casuscelli) wrote in message news:<8a8f14b9.0402031647.5e035d09@posting.google.com>... > > How many tables are you going to replicate? > > Schema changes will affect you production enviroment, ER is based on > > the schema for each table. > > Also you will have a lot of administration for an update anywhere env. > > rather than HDR. > > a big number of logical logs if it is a high txn volume. > > > > esteban.- > > > > mosserp@wellsfargo.com wrote in message news:<bvoo8m$guc$1@terabinaries.xmission.com>... > > > We have been told to figure out how to replace our current HDR environment > > > with an ER update-anywhere solution, in order to eliminate "outages" > > > (planned or unplanned) with the DB server. > > > > > > Our environment: > > > -- HP-UX 11.0 on identical HP RP7400 servers (8 cpu at 750 MHz, 16Gb memory) > > > -- IDS 9.30.FC4XC > > > -- servers are located hundreds of miles from each other > > > -- high txn volume > > > -- HA is *extremely* critical, no regular maintenance window > > > -- schema changes occur 2-4 times per year > > > -- for planned maintenance, we typically need two 5-10 minute outages to > > > switch secondary to standard mode and move the client connections over -- > > > client wants these outages, plus all unplanned outages, eliminated > > > completely - figures that ER is the answer > > > > > > Has anyone else implemented ER in update-anywhere (aka, peer-to-peer) mode > > > in a high-volume, extremely high-availability environment? If so, what > > > should we watch out for, etc.? > > > > > > Comments? Recommendations? > > > > > > TIA, > > > Paul Mosser > > > sending to informix-list > > > sending to informix-list
scottishpoet wrote: > But seriously, I'm not sure what the client will see here that they > don't see at the moment with HDR. Maybe Madisson will shoot me down in > flames, but ER won't redirect network connections from one server to > other other when the server goes down any more than HDR will. You will > still need to stop communication with their server and redirect the > connections to the other server. So the impact will be reduced, but > not eliminated! It's simple if the application is aware that there are multiple potential data sources and can fail over between them seamlessly. -- "C'est pas parce qu'on n'a rien ''' dire qu'il faut fermer sa gueule" - Coluche