You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@zookeeper.apache.org by "Robert Stone (Jira)" <ji...@apache.org> on 2022/07/15 17:12:00 UTC
[jira] [Comment Edited] (ZOOKEEPER-1622) session ids will be negative in the year 2022
[ https://issues.apache.org/jira/browse/ZOOKEEPER-1622?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17567214#comment-17567214 ]
Robert Stone edited comment on ZOOKEEPER-1622 at 7/15/22 5:11 PM:
------------------------------------------------------------------
I'm running 3.4.5 (from Cloudera) on a few systems. On two of them I recently restarted ZK and now Accumulo is not happy, complaining about NumberFormatException: For input string: "ff81f7...." Is this the same thing? How do I actually tell if the session id is negative?
This WAS my issue, and it was resolved with the described fix.
was (Author: JIRAUSER292897):
I'm running 3.4.5 (from Cloudera) on a few systems. On two of them I recently restarted ZK and now Accumulo is not happy, complaining about NumberFormatException: For input string: "ff81f7...." Is this the same thing? How do I actually tell if the session id is negative?
> session ids will be negative in the year 2022
> ---------------------------------------------
>
> Key: ZOOKEEPER-1622
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1622
> Project: ZooKeeper
> Issue Type: Bug
> Affects Versions: 3.4.0, 3.5.0
> Reporter: Eric C. Newton
> Assignee: Eric C. Newton
> Priority: Trivial
> Fix For: 3.4.6, 3.5.0
>
> Attachments: ZOOKEEPER-1622.patch
>
>
> Someone decided to use a large number for their myid file. This cause session ids to go negative, and our software (Apache Accumulo) did not handle this very well. While diagnosing the problem, I noticed this in SessionImpl:
> {noformat}
> public static long initializeNextSession(long id) {
> long nextSid = 0;
> nextSid = (System.currentTimeMillis() << 24) >> 8;
> nextSid = nextSid | (id <<56);
> return nextSid;
> }
> {noformat}
> When the 40th bit in System.currentTimeMillis() is a one, sign extension will fill the upper 8 bytes of nextSid, and id will not make the session id unique. I recommend changing the right shift to the logical shift:
> {noformat}
> public static long initializeNextSession(long id) {
> long nextSid = 0;
> nextSid = (System.currentTimeMillis() << 24) >>> 8;
> nextSid = nextSid | (id <<56);
> return nextSid;
> }
> {noformat}
> But, we have until the year 2022 before we have to worry about it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)