liuml07 commented on a change in pull request #2189:
URL: https://github.com/apache/hadoop/pull/2189#discussion_r483414755
##########
File path:
hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/StorageType.java
##########
@@ -34,28 +34,35 @@
@InterfaceStability.Unstable
public enum StorageType {
// sorted by the speed of the storage types, from fast to slow
- RAM_DISK(true),
- SSD(false),
- DISK(false),
- ARCHIVE(false),
- PROVIDED(false);
+ RAM_DISK(true, true),
+ NVDIMM(false, true),
+ SSD(false, false),
+ DISK(false, false),
+ ARCHIVE(false, false),
+ PROVIDED(false, false);
private final boolean isTransient;
+ private final boolean isRAM;
public static final StorageType DEFAULT = DISK;
public static final StorageType[] EMPTY_ARRAY = {};
private static final StorageType[] VALUES = values();
- StorageType(boolean isTransient) {
+ StorageType(boolean isTransient, boolean isRAM) {
this.isTransient = isTransient;
+ this.isRAM = isRAM;
}
public boolean isTransient() {
return isTransient;
}
+ public boolean isRAM() {
+ return isRAM;
+ }
Review comment:
> neither RAM nor nvdimm need FsDatasetCache, and isTransient() used to
determine whether FsDatasetCache is needed
I think we have agreed on this.
> I think it is better to use isRAM to determine whether to use mover
So just to be clear, if we use `isRAM()` to determine, both RAM_DISK and
NVDIMM will return true and thus will simply disable Balancer and Mover. I was
proposing that only when we do not want to move those NVDIMM data replicas
around (disk/node) ever. Is that the design here? I'm fine with that but I
think this contradicts with the statement that _"It can be regarded as a
general hardware device. "_
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
