Fix the bug after introducing ufs full path as page id

[YichuanSun]

In Alluxio 3.3, we use the ufs full path as the fileId, instead of a hash String. It breaks the fsspec/alluxiofs, this PR is used to fix the bugs.

[YichuanSun]

I have run the fixed alluxiofs with the Alluxio 3.3 locally, it passed the test_read_range.py. @LuQQiu could you please take a look?

[LuQQiu]

@SibylYang please help review and see whether that will break backward compatibility? e.g. whether it can still work with AOS

[YichuanSun]

I believe that the "fake server" in test_fake_server.py is not compatible with modified code. But I'm not sure how to fix the fake server. By the way, test_fake_server.py is the only failed test.

[SibylYang]

I tested it with alluxio open source cluster running locally. There's no problem with reading files. But there's an error when writing to files. Below code generated the error "io.UnsupportedOperation: not writable".

`
fsspec.register_implementation("alluxiofs", AlluxioFileSystem, clobber=True)
alluxio_fs = fsspec.filesystem("alluxiofs", etcd_hosts="localhost", etcd_port=2379, target_protocol='s3')

with alluxio_fs.open('s3://path/to/file', 'a') as f:
f.write("\n new line by alluxio\n")
`

[YichuanSun]

After using the code in #51 and my change in current PR, I test the write function manually with the latest alluxio os edition, I find the write works. Here is my test script, the same as yours above.

import sys
import fsspec
sys.path.append('.')
from alluxiofs import AlluxioClient, AlluxioFileSystem

if __name__ == "__main__":
    fsspec.register_implementation("alluxiofs", AlluxioFileSystem, clobber=True)
    alluxio_fs = fsspec.filesystem("alluxiofs", etcd_hosts="localhost", etcd_port=2379, target_protocol='s3')
    with alluxio_fs.open("s3://yichuan-test/testBucket/test.txt", 'a') as f:
        f.write("\n444 new line by alluxio\n")

After running the script and open the "s3://yichuan-test/testBucket/test.txt" in the S3 website, the content is as below.

Wow

 new line by alluxio

 new line by alluxio

 new line by alluxio

111 new line by alluxio

222 new line by alluxio

333 new line by alluxio

444 new line by alluxio

Maybe just any wrong configurations that cause the write error? FYI @LuQQiu @SibylYang

[SibylYang]

After using the code in #51 and my change in current PR, I test the write function manually with the latest alluxio os edition, I find the write works. Here is my test script, the same as yours above.

import sys
import fsspec
sys.path.append('.')
from alluxiofs import AlluxioClient, AlluxioFileSystem

if __name__ == "__main__":
    fsspec.register_implementation("alluxiofs", AlluxioFileSystem, clobber=True)
    alluxio_fs = fsspec.filesystem("alluxiofs", etcd_hosts="localhost", etcd_port=2379, target_protocol='s3')
    with alluxio_fs.open("s3://yichuan-test/testBucket/test.txt", 'a') as f:
        f.write("\n444 new line by alluxio\n")

After running the script and open the "s3://yichuan-test/testBucket/test.txt" in the S3 website, the content is as below.

Wow

 new line by alluxio

 new line by alluxio

 new line by alluxio

111 new line by alluxio

222 new line by alluxio

333 new line by alluxio

444 new line by alluxio

Maybe just any wrong configurations that cause the write error? FYI @LuQQiu @SibylYang

Right. If I used PR 51 --- the updated handler, then there's no problem with writing

[LuQQiu]

@SibylYang Please help review the PR

[SibylYang]

This commit LGTM.
It won't break the alluxio open source, because in our alluxio java server, it process URL using something like
httpRequestUri.getParameters().get("offset")
So the added "ufsFullPath" in the URL will just be ignored by alluxio open source.

[LuQQiu]

LGTM