[infinispan-issues] [JBoss JIRA] Issue Comment Edited: (ISPN-1275) Virtual nodes producing same hash id for two different cluster nodes

[ https://issues.jboss.org/browse/ISPN-1275?page=com.atlassian.jira.plugin.... ] Dan Berindei edited comment on ISPN-1275 at 7/29/11 6:38 AM: ------------------------------------------------------------- Manik, it's a general consistent hash thing, if there is a collision (two nodes get the same normalized hash) then the second node joining or leaving has a 50% chance of changing the other node's hash id. Obviously with virtual nodes the chance of a collision is much higher, and with 2000 nodes in a 10240 hash space it's really close to 100%. I think this also opens the possibility for ISPN-1170 to happen again: {code} hash(k) = 101 ch_0 = {A{r0,m0} = 100, B{r0,m0} = 102, C{r0,m1} = 103} loc_0(k) = {B, C} //after D{r1,m0} == 100 joins ch_1 = {D{r1,m0} = 100, A{r0,m0} = 101, B{r0,m0} = 102, C{r0,m1} = 103} loc_1(k) = {A, D} {code} was (Author: dan.berindei): Manik, it's a general consistent hash thing, if there is a collision (two nodes get the same normalized hash) then the second node joining or leaving has a 50% chance of changing the other node's hash id. Obviously with virtual nodes the chance of a collision is much higher, and with 2000 nodes in a 10240 hash space it's really close to 100%. I think this also opens the possibility for ISPN-1170 to happen again: {code} hash(k) = 101 ch_0 = {A{r0,m0} = 100, B{r0,m0} = 102, C{r0,m1} = 103} loc_0(k) = {B, C} //after D{r1,m0} == 100 joins ch_0 = {D{r1,m0} = 100, A{r0,m0} = 101, B{r0,m0} = 102, C{r0,m1} = 103} loc_1(k) = {A, D} {code} -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira