fix(downloads): do not download into browser memory #319
Conversation
|
Apparently I can't remove my request for review @jan-law |
hmm, me neither |
andrewazores
force-pushed
the
download-flow
branch
3 times, most recently
from
1a7c547 to
4e41bc8
Compare
Previously, the file download flow (for templates, reports, and even binary recording files(!)) would download the remote file into the browser's memory, then create a local objectURL for the blob, assign that URL to an anchor element, and prompt the user to "download" the asset from that objectURL. This means that the user's "Cancel" selection was somewhat meaningless - by the time they see this, the data has already been fully transferred to their machine, and all they are cancelling is to write it to disk. Worse, this means that the time to display the dialog depends on the time to transfer the data to the user's machine. For a remote OpenShift cluster sending a large JFR binary file, this time could become very significant. This commit improves the download handling by instead relying on the HTML5 "download" attribute of an anchor element to instruct the browser to download the asset linked by the href, rather than attempt to open it, and to use a web-client-defined name for the downloaded file (note that browsers do not respect this attribute for cross-origin requests, so some file downloads may be badly named in webpack-dev-server setups). This vastly decreases the time it takes for the browser's Save File dialog to appear, and allows the user to cancel the action and prevent the data from being transferred to begin with.
andrewazores
force-pushed
the
download-flow
branch
from
4e41bc8 to
e4064e4
Compare
|
@jan-law this is ready for re-review. This is dependent on a large backend PR as well, but this PR on its own can be reviewed mostly independently. If some backend changes are made during the review they shouldn't change the general concept and download flow too much, just perhaps the API paths or response formats. |
There was a problem hiding this comment.
Not sure if this comment applies to the frontend or backend - when I try to archive an active, running recording from cryostat:9094 with Basic authentication enabled, I see an AccessDeniedException
Nov 12, 2021 6:06:16 PM io.cryostat.core.log.Logger info
INFO: (10.0.2.100:59474): OPTIONS /api/v1/targets/service%3Ajmx%3Armi%3A%2F%2F%2Fjndi%2Frmi%3A%2F%2Fcryostat%3A9094%2Fjmxrmi/recordings/test 200 0ms
Nov 12, 2021 6:06:16 PM io.cryostat.core.log.Logger error
SEVERE: HTTP 500: java.nio.file.AccessDeniedException: /opt/cryostat.d/recordings.d/ONSXE5TJMNSTU2TNPA5HE3LJHIXS6L3KNZSGSL3SNVUTULZPMNZHS33TORQXIORZGA4TIL3KNV4HE3LJ
io.vertx.ext.web.handler.impl.HttpStatusException: Internal Server Error
Caused by: java.util.concurrent.ExecutionException: java.nio.file.AccessDeniedException: /opt/cryostat.d/recordings.d/ONSXE5TJMNSTU2TNPA5HE3LJHIXS6L3KNZSGSL3SNVUTULZPMNZHS33TORQXIORZGA4TIL3KNV4HE3LJ
at java.base/java.util.concurrent.CompletableFuture.reportGet(CompletableFuture.java:395)
at java.base/java.util.concurrent.CompletableFuture.get(CompletableFuture.java:1999)
at io.cryostat.net.web.http.api.v1.TargetRecordingPatchSave.handle(TargetRecordingPatchSave.java:67)
at io.cryostat.net.web.http.api.v1.TargetRecordingPatchHandler.handleAuthenticated(TargetRecordingPatchHandler.java:108)
at io.cryostat.net.web.http.AbstractAuthenticatedRequestHandler.handle(AbstractAuthenticatedRequestHandler.java:90)
at io.cryostat.net.web.http.AbstractAuthenticatedRequestHandler.handle(AbstractAuthenticatedRequestHandler.java:65)
at io.vertx.ext.web.impl.BlockingHandlerDecorator.lambda$handle$0(BlockingHandlerDecorator.java:48)
at io.vertx.core.impl.ContextImpl.lambda$executeBlocking$2(ContextImpl.java:313)
at io.vertx.core.impl.TaskQueue.run(TaskQueue.java:76)
at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.base/java.lang.Thread.run(Thread.java:829)
Caused by: java.nio.file.AccessDeniedException: /opt/cryostat.d/recordings.d/ONSXE5TJMNSTU2TNPA5HE3LJHIXS6L3KNZSGSL3SNVUTULZPMNZHS33TORQXIORZGA4TIL3KNV4HE3LJ
at java.base/sun.nio.fs.UnixException.translateToIOException(UnixException.java:90)
at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:111)
at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:116)
at java.base/sun.nio.fs.UnixFileSystemProvider.createDirectory(UnixFileSystemProvider.java:389)
at java.base/java.nio.file.Files.createDirectory(Files.java:690)
at io.cryostat.recordings.RecordingArchiveHelper.writeRecordingToDestination(RecordingArchiveHelper.java:306)
at io.cryostat.recordings.RecordingArchiveHelper.lambda$saveRecording$0(RecordingArchiveHelper.java:138)
at io.cryostat.net.TargetConnectionManager.executeConnectedTask(TargetConnectionManager.java:136)
at io.cryostat.recordings.RecordingArchiveHelper.saveRecording(RecordingArchiveHelper.java:131)
... 11 more
Nov 12, 2021 6:06:16 PM io.cryostat.core.log.Logger info
INFO: (10.0.2.100:59478): PATCH /api/v1/targets/service%3Ajmx%3Armi%3A%2F%2F%2Fjndi%2Frmi%3A%2F%2Fcryostat%3A9094%2Fjmxrmi/recordings/test 500 40ms
This is a backend problem. Are you using |
I'm using |
Okay, so in that case the |
|
@jan-law check https://github.com/cryostatio/cryostat/issues/751 , I included a diff there that you can apply to your |
I observed this too and thought it was kind of a nice change - it lets you view the document in "full screen" (not in a miniature iframe anyway). You can still do a typical I think I can instruct the browser to actually download it rather than viewing it in a new tab by setting the According to the MDN docs on |
I like the idea of a full screen report with a "View Report" action. |
Fixes #318
Related to cryostatio/cryostat#719
Previously, the file download flow (for templates, reports, and even
binary recording files(!)) would download the remote file into the
browser's memory, then create a local objectURL for the blob, assign
that URL to an anchor element, and prompt the user to "download" the
asset from that objectURL. This means that the user's "Cancel" selection
was somewhat meaningless - by the time they see this, the data has already
been fully transferred to their machine, and all they are cancelling is
to write it to disk. Worse, this means that the time to display the
dialog depends on the time to transfer the data to the user's machine.
For a remote OpenShift cluster sending a large JFR binary file, this
time could become very significant.
This commit improves the download handling by instead relying on the
HTML5 "download" attribute of an anchor element to instruct the browser
to download the asset linked by the href, rather than attempt to open
it, and to use a web-client-defined name for the downloaded file (note
that browsers do not respect this attribute for cross-origin requests,
so some file downloads may be badly named in webpack-dev-server setups).
This vastly decreases the time it takes for the browser's Save File
dialog to appear, and allows the user to cancel the action and prevent
the data from being transferred to begin with.