[2.0.1 RN] More 2.0.1 changes in cbepctl

Karen Zeller added a comment - 13/Feb/13 6:52 PM Hi Jin, Can you clarify: In the past you could not update these settings while a node is online and now you can? Please also provide three examples including bucket info in needed for the new parameters. Thanks!

Mike Wiederhold added a comment - 15/Feb/13 3:43 PM mutation_mem_threshold is the only new stat and works the same as the other two which should already be in your documentation. Can you check to see if you already have mem_high_wat and mem_low_wat documented. If not please assign this to me and I will provide examples.

Mike Wiederhold added a comment - 15/Feb/13 3:45 PM mutation_mem_threshold is the amount of memory that can be used before the client starts receiving the temp oom message.

Jin Lim added a comment - 15/Feb/13 3:57 PM Thanks Mike for detailed explanation. ex: password is only required for SASL authenticating bucket and default bucket name is "default" ./cbepctl 10.5.2.31:11210 -b bucket_name -p password set flush_param mem_high_wat 70 ./cbepctl 10.5.2.31:11210 -b bucket_name -p password set flush_param mem_low_wat 60 ./cbepctl 10.5.2.31:11210 -b bucket_name -p password set flush_param mutation_mem_threshold 65

Karen Zeller added a comment - 04/Mar/13 4:03 PM I've already got the high water and low water documented. For this new one mutation_mem_threshold, do you have guideance on when customers should change this and when not to change this? Also, does changing this setting impact other parts of server operation or client/server ops that people should be aware of (e.g warmup) What will they observe and should observe if they change it? Please add and assign back. Karen

Karen Zeller added a comment - 04/Mar/13 4:04 PM P.S. I'm thinking this impacts ejection, yes?

Mike Wiederhold added a comment - 04/Mar/13 6:27 PM This impacts when you start receiving tmp oom messages on in your client application. By default this value is now when memory is 95% full and this leaves 5% of your memory for overhead. I would only recommend people changing this value (higher) if they have a lot of memory or (lower) if they think they need more space for the overhead, like data structures and disk queues, of Couchbase.

Karen Zeller added a comment - 05/Mar/13 2:33 PM Ok Added to RN 2.0.1: There is a new setting available in <command>cbepctl</command> named <command>mutation_mem_threshold</command>. This is the amount of RAM that can be used on the server before a client begins receiving temporary out of memory errors. For more information, see <xref linkend="couchbase-admin-cbepctl-mutation_mem" />. And then cross ref to new section: Changing Setting for Out Of Memory Errors By default, Couchbase Server will send clients a temporary out of memory error if RAM is 95% consumed and only 5% RAM remains for overhead. We do not suggest you change this default to a higher value; however you may choose to reduce this value if you think you need more RAM available for system overhead such as disk queue or for server data structures. To change this value: shell>./cbepctl 10.5.2.31:11210 -b bucket_name -p password set flush_param mutation_mem_threshold 65 In this example we reduce the threshold to 65% of RAM. This setting must be updated on a per-node, per-bucket basis, meaning you need to provide the specific node and named bucket to update this setting. To update it for an entire cluster, you will need to issue the command for every combination of node and named bucket that exists in the cluster.

Karen Zeller added a comment - 05/Mar/13 2:34 PM Ok Added to RN 2.0.1: There is a new setting available in <command>cbepctl</command> named <command>mutation_mem_threshold</command>. This is the amount of RAM that can be used on the server before a client begins receiving temporary out of memory errors. For more information, see <xref linkend="couchbase-admin-cbepctl-mutation_mem" />. And then cross ref to new section: Changing Setting for Out Of Memory Errors By default, Couchbase Server will send clients a temporary out of memory error if RAM is 95% consumed and only 5% RAM remains for overhead. We do not suggest you change this default to a higher value; however you may choose to reduce this value if you think you need more RAM available for system overhead such as disk queue or for server data structures. To change this value: shell>./cbepctl 10.5.2.31:11210 -b bucket_name -p password set flush_param mutation_mem_threshold 65 In this example we reduce the threshold to 65% of RAM. This setting must be updated on a per-node, per-bucket basis, meaning you need to provide the specific node and named bucket to update this setting. To update it for an entire cluster, you will need to issue the command for every combination of node and named bucket that exists in the cluster.