Bring feature/perf up-to-date with master #5582
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This is about the same order of magnitude as the number of VMs supported in a host Signed-off-by: Pau Ruiz Safont <[email protected]>
This helps when running it to observe behaviour of the library Signed-off-by: Pau Ruiz Safont <[email protected]>
Add standard http attributes to from opentelemetry such as: - `http.request.method` - `http.request.header.content-type`; - `http.request.body.size`; - `http.request.header.user-agent`; - `http.request.header.(key)`. This improves debuggability associated with client requests. Signed-off-by: Gabriel Buica <[email protected]>
This reuses the re-key process to gather current contents of the cache and depending on the amount of elements pending it 1. Bypasses the cache completely if it's empty 2. Uses the queue if all the elements fit in the queue 3. Starts as before if there are more elements Signed-off-by: Pau Ruiz Safont <[email protected]>
Unknown directories detected in the cache will get logged, and at the appropriate times deleted, along with all the files contained within. Signed-off-by: Pau Ruiz Safont <[email protected]>
Previously a logline was created per push attempt, now up to two per failure period are done: one on the first failure, and one on recovery. Signed-off-by: Pau Ruiz Safont <[email protected]>
Now temporary files are an internal detail of the persistence function. The rest of code needs not be aware of it: if some of temp files linger in the filesystem means that the process did not acknowledge a requested write, so the domain could not have possibly acted on the write. Delete these as this does not incur in any data corruption in the domain side. Signed-off-by: Pau Ruiz Safont <[email protected]>
With *hybrid source format in export, the following cases are supported: 1. nbdhybrid: QCOW2 -> NBD device in dom0 -> exported file 2. nbdhybrid: VHD -> NBD device in dom0 -> exported file 3. hybrid: VHD -> blktap device in dom0 -> exported file The case 2 above can't support an optional parameter "base". This parameter holds the ID of another VHD VDI. When it is passed, the export will only write the differences between "source" and "base" to the destination file. As a short-term solution, in case 2 above, when the "base" is passed, the source format is changed to "hybrid" in this commit. This can work because: 1. the comparsion on blocks required by "base" is supported by "hybrid"; 2. the raw data from NBD device and blktap (Frankentap) device are same; 3. the sparseness information of the source required in case 2 can be get by either NBD interface (nbdhybrid) or VHD parsing (hybrid). Signed-off-by: Ming Lu <[email protected]>
waitpid_nohang closes the socket by calling waitpid and then calls clear_nonblock on that socket, which fails. waitpid_nohang's policy around closing the socket is not obvious enough. Make it more explicit by not calling waitpid but inlining code and looking at the different cases. The behaviour is still to close the socket when the child process has terminated and not otherwise. Logging can be re-enabled if we have to come back to this. Signed-off-by: Christian Lindig <[email protected]>
…384483 CA-384483: Can't export VDI to VHD file with base VDI
Signed-off-by: Lunfan Zhang <[email protected]>
Signed-off-by: Benjamin Reis <[email protected]>
Slightly different from missing file, add missing path. Signed-off-by: Frediano Ziglio <[email protected]>
Clean even if user press Ctrl-C, not only sending TERM signal. Remove possible stale unix socket to allow to run multiple times. Signed-off-by: Frediano Ziglio <[email protected]>
Make sure we get the correct output from program launched. Signed-off-by: Frediano Ziglio <[email protected]>
Make sure program receives the string we are sending. Signed-off-by: Frediano Ziglio <[email protected]>
Requires libxml2-devel installed. Signed-off-by: Edwin Török <[email protected]> (cherry picked from commit 0008395)
The C SDK build was failing with a recent GCC on Fedora39 like this:
```
src/xen_common.c: In function ‘xen_session_logout’:
src/xen_common.c:298:5: error: ‘xen_call_’ accessing 16 bytes in a region of size 0 [-Werror=stringop-overflow=]
298 | xen_call_(session, "session.logout", params, 0, NULL, NULL);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
Use `NULL` instead of a VLA of size 0.
Signed-off-by: Edwin Török <[email protected]>
(cherry picked from commit bfe78bd)
…dora-fix Port of "Fix C SDK build on Fedora39" to `master`
Domains using TPMs need to have SWTPM always available. Encode this information in the service file so the services are shut down at the correct time while powering off the hosts. Signed-off-by: Pau Ruiz Safont <[email protected]>
I am not sure where this restriction came from. It was introduced in the same commit that introduced the metadata-export feature itself, without explanation. Full exports of snapshots are already allowed. Signed-off-by: Rob Hoes <[email protected]>
This allows clients to know which snapshots they knowingly just created and act on them instead of looking again into the db and try to divine which snapshots the call created. Signed-off-by: Pau Ruiz Safont <[email protected]>
This allows clients to detect that the newer device type exclusions for export are available. Signed-off-by: Rob Hoes <[email protected]>
This allows to orchestrators to copy devices like VBDs from the VM, export the metadata without the disk, import the VM to any pool, and finally add the saved VBDs to the new VM before booting it. Signed-off-by: Pau Ruiz Safont <[email protected]>
This allows to easily test the http endpoint. Also advertises the metadata parameter on these calls Signed-off-by: Pau Ruiz Safont <[email protected]>
Use the helper functions to process parameters, use atomics instead of ad-hoc references for counting in possibly-parallel case Signed-off-by: Pau Ruiz Safont <[email protected]>
Signed-off-by: Pau Ruiz Safont <[email protected]>
Signed-off-by: Pau Ruiz Safont <[email protected]>
Signed-off-by: Pau Ruiz Safont <[email protected]>
…dk-c CA-387885 and templatization of the C SDK
Attempts to split the tracing library into components and exporter parts. While trying to instrument `forkhelpers.ml` with tracing, I found a cycle dependency: `Tracing(Exporter)` -> `Open_uri` -> `Stunnel` -> `Forkhelpers`. This splits the tracing library between exporting functionality and components. Signed-off-by: Gabriel Buica <[email protected]>
Small improvements to `tracing_export.ml` and `tracing_exoprt.mli`. Issues found while splitting the library such as: missing documentation and small code refactoring for better readability. Signed-off-by: Gabriel Buica <[email protected]>
Adds `with_tracing` helper function to `forkhelpers.ml`. This is to show that there are no more cycle dependecies between `tracing` library and `forexecd` and it will be used in a follow-up PR to instrument `forkhelpers` with tracing. Signed-off-by: Gabriel Buica <[email protected]>
As discussed here: xapi-project/xapi-project.github.io#286 New methods: - `PCI.disable_dom0_access`: Hide a PCI from dom0 kernel - `PCI.enable_dom0_access`: Unhide a PCI from dom0 kernel - `PCI.get_dom0_access_status`: Return a PCI device dom0 access status The answer is no longer based on a DB field but by comparing the `/proc/cmdline` and the xen cmdline. Deprecated methods: - `PGPU.disable_dom0_acces` - `PGPU.enable_dom0_access` The methods should be called on the PCI belonging to the PGPU instead Deprecated field: - `PGPU.dom0_access` The deprecated methods still work since they now call the PCI methods. The PGPU field is kept up to date with the statuss returned by the PCI methods. Corresponding xe CLI calls have been implemented. Signed-off-by: Benjamin Reis <[email protected]>
It was allocating a string for each character, only to then immediately add it to a buffer. Instead add the character to the buffer directly. The temporary string was probably created to factor out 'Buffer.add', but I'd rather copy Buffer.add 4x times than have the perf hit. Further optimizations might be possible here as the comments says, but this is a low-risk, obvious optimization. Signed-off-by: Edwin Török <[email protected]>
Optimistically check for the presence of escape characters before escaping. This avoids rebuilding the string when escapable characters are not present (the common case). Signed-off-by: Edwin Török <[email protected]>
Check for the presence of the escape character before rebuilding the string Signed-off-by: Edwin Török <[email protected]>
…expr
IH-553: Optimize Sexpr.{escape,unescape}
…/CP-48195-v2 CP-48195: Split `tracing` library to avoid future cyclic dependencies
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
|
xe-xentrace was changed recently, will look into fixing it so it passes the CI. |
|
Once I fixed some of the warnings shellcheck did actually find this one that points to an actual bug: |
Fix shellcheck warnings and bugs. Signed-off-by: Edwin Török <[email protected]>
edwintorok
force-pushed
the
feature/perf
branch
from
51458f5
to
916b6e0
Compare
psafont
approved these changes
Apr 22, 2024
psafont
requested changes
Apr 22, 2024
Good spot, fixed |
psafont
approved these changes
Apr 22, 2024
Vincent-lau
approved these changes
Apr 23, 2024
Vincent-lau
reviewed
Apr 23, 2024
|
"edwintorok wants to merge 80 commits into xapi-project:feature/perf from edwintorok:feature/perf" I think it merged into the correct place into xapi-project What we do here is merge upstream master into upstream feature/perf, but there are conflicts that I need to fix, so that can only be done from my private fork (I can't directly push to feature/perf on upstream) |
pytype_reporter extracted 47 problem reports from pytype output. You can check the results of the job here |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixed some conflicts in the Makefile, C bindings gen, and a build issue with pgpu_dom0_access_to_string