> On May 13, 2014, 9:55 p.m., Robert Kanter wrote: > > ID's usually look like this: 0000000-140505141334644-oozie-oozi-W > > where the second value is a timestamp and the first value is the increasing > > part. > > > > With the ZKUUIDService, will all Oozie servers use the same timestamp part? > > If not, then having the first part shared among the servers doesn't get > > you too much; so I think we need to have a way of making the timestamp > > shared as well. > > > > Also, when you reset the sequence, isn't it possible to still cause > > collisions if someone has a really old job still sitting around? I agree > > that it's probably not likely given the number of jobs they'd have to run > > without cleaning up, but still. If you also reset the timestamp to "now" > > when reseting the atomic long, that should take care of this.
>With the ZKUUIDService, will all Oozie servers use the same timestamp part? >If not, then having the first part shared among the servers doesn't get you >too much; so >I think we need to have a way of making the timestamp shared as >well. In HA Timestamp may tell, where job was submitted. >Also, when you reset the sequence, isn't it possible to still cause collisions >if someone has a really old job still sitting around? I agree that it's >probably not >likely given the number of jobs they'd have to run without >cleaning up, but still. If you also reset the timestamp to "now" when >reseting the atomic long, that >should take care of this. Once we restart, timestamp will change. If one knows that there are huge jobs submission and less purging, they can increase max count. Reseting timestamp to "now" might take out the purpose of appending timestamp to jobid. I believe taking out timestamp for HA is debatable. I will create a JIRA and if all agrees we can take it out. > On May 13, 2014, 9:55 p.m., Robert Kanter wrote: > > core/src/main/java/org/apache/oozie/service/ZKUUIDService.java, line 55 > > <https://reviews.apache.org/r/21046/diff/3/?file=580589#file580589line55> > > > > I'm not sure we should make the path configurable; I'm concerned that > > will just lead to a user accidentally setting them different in different > > servers and then running into problems because Oozie servers are using > > different paths. Or what happens if they try to use a patch Oozie is > > already using for something else? > > > > Also, please verify that when you run Oozie with ZKUUIDService, the > > path gets created under the namespace set for the server (default "oozie") > > and not simply at the root. e.g. /oozie/job_id_sequence > > > > Also, can you add a class javadoc comment to ZKUUIDService mentioning > > the path it's using (see ZKUtils and ZKLocksService for an example of what > > I mean) >I'm not sure we should make the path configurable; I'm concerned that will >just lead to a user accidentally setting them different in different servers >and then >running into problems because Oozie servers are using different >paths. Or what happens if they try to use a patch Oozie is already using for >something else? Even Namespace is configurable, it should be ok. Most of the people will verify their setting before moving to production. I have verified and our QE will also verify that. Path are created under namespace. > On May 13, 2014, 9:55 p.m., Robert Kanter wrote: > > core/src/main/java/org/apache/oozie/service/ZKUUIDService.java, line 90 > > <https://reviews.apache.org/r/21046/diff/3/?file=580589#file580589line90> > > > > I think you mean "increment" not "reset" here? I mean reset, we try to reset three time, before throwing exception. - Purshotam ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/21046/#review42891 ----------------------------------------------------------- On May 13, 2014, 5:32 p.m., Purshotam Shah wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/21046/ > ----------------------------------------------------------- > > (Updated May 13, 2014, 5:32 p.m.) > > > Review request for oozie. > > > Bugs: OOZIE-1715 > https://issues.apache.org/jira/browse/OOZIE-1715 > > > Repository: oozie-git > > > Description > ------- > > OOZIE-1715 Distributed ID sequence for HA > > > Diffs > ----- > > core/src/main/java/org/apache/oozie/service/DistributedUUIDService.java > e69de29 > core/src/main/java/org/apache/oozie/service/UUIDService.java 836815d > core/src/main/java/org/apache/oozie/service/ZKUUIDService.java e69de29 > core/src/main/java/org/apache/oozie/util/ZKUtils.java 56055b8 > core/src/test/java/org/apache/oozie/service/TestDistributedUUIDService.java > e69de29 > core/src/test/java/org/apache/oozie/service/TestZKUUIDService.java e69de29 > > Diff: https://reviews.apache.org/r/21046/diff/ > > > Testing > ------- > > UTC > > > Thanks, > > Purshotam Shah > >
