What should be the minimum for creating an ID that would have chances of being unique among containers?
And what is important within a distributed system for an ID?
What about the below:
a. ID is of type uint64 with maximum being 18446744073709551615 (20 digits).
b. epoch time
c. container ID
d. some random number
e. some sequence
Why not process ID? Due to the fact that a container should only run one process.
A value up to a 1/10 of a second was chosen for good precision.
So 1654589015 is seconds, length is 10 digits. Our ID would use 11 digits.
This gives us another 9 positions.
The value would be read from /etc/machine-id.
Would cover 5 postions.
Would cover 4 positions.
The value would be compensated with the epoch time nano second positions to
always reach the length.
The constructor to be used is NewIDRandom.
Would cover 4 positions as it replaces the random number in NewIDIncremental10K.
This constructor should be used with traffic up to 10000 requests per 1/10 of a second. It is faster than the random one as per the benchmarks and was tested for race conditions.
Go routines do not spawn new process IDs as per:
package main
import (
"fmt"
"os"
"sync"
)
func main() {
res := make(map[int]struct{})
var wg sync.WaitGroup
pid := func() {
p := os.Getpid()
if _, exists := res[p]; !exists {
res[p] = struct{}{}
}
wg.Done()
}
no := 50
wg.Add(no)
for i := 0; i < no; i++ {
go pid()
}
wg.Wait()
fmt.Println(res)
}https://www.sobyte.net/post/2022-01/go-snowflakeid/