[
https://issues.apache.org/jira/browse/ZOOKEEPER-1616?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13547589#comment-13547589
]
liang xie commented on ZOOKEEPER-1616:
--------------------------------------
Cooooooool, [~tlipcon]
One thing need to explain is that the implement of currentTimeMillis&nanoTime
is affected by lower OS version as well，so the good proposal will benefit newer
OS only. Here is the code snappiet from hs20 just for reference :
{code}
jlong os::javaTimeNanos() {
if (Linux::supports_monotonic_clock()) {
struct timespec tp;
int status = Linux::clock_gettime(CLOCK_MONOTONIC, &tp);
assert(status == 0, "gettime error");
jlong result = jlong(tp.tv_sec) * (1000 * 1000 * 1000) + jlong(tp.tv_nsec);
return result;
} else {
timeval time;
int status = gettimeofday(&time, NULL);
assert(status != -1, "linux error");
jlong usecs = jlong(time.tv_sec) * (1000 * 1000) + jlong(time.tv_usec);
return 1000 * usecs;
}
}
jlong os::javaTimeMillis() {
timeval time;
int status = gettimeofday(&time, NULL);
assert(status != -1, "linux error");
return jlong(time.tv_sec) * 1000 + jlong(time.tv_usec / 1000);
}
{code}
> time calculations should use a monotonic clock
> ----------------------------------------------
>
> Key: ZOOKEEPER-1616
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1616
> Project: ZooKeeper
> Issue Type: Bug
> Reporter: Todd Lipcon
>
> We recently had an issue with ZooKeeper sessions acting strangely due to a
> bad NTP setup on a set of hosts. Looking at the code, ZK seems to use
> System.currentTimeMillis to measure durations or intervals in many places.
> This is bad since that time can move backwards or skip ahead by several
> minutes. Instead, it should use System.nanoTime (or a wrapper such as Guava's
> Stopwatch class)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
