added a new admin endpoint - /logs to fetch log files

[akash-a-n]

• Implement GET /logs?type=<mlog|qlog>&connection_id=<hex> route.
• Offload blocking file I/O (std::ifstream::read) to folly::getGlobalCPUExecutor() to avoid stalling the admin event loop.

---

This change is 

[afrind]

@afrind reviewed 8 files and all commit messages, and made 6 comments.
Reviewable status: all files reviewed, 6 unresolved discussions (waiting on akash-a-n).

---

src/admin/ConnectionLogsHandler.cpp line 54 at r1 (raw file):

// Read an entire file into an IOBuf. Returns nullptr if the file cannot be
// opened, is empty, or exceeds maxBytes.
std::unique_ptr<folly::IOBuf> readFileToIOBuf(const std::string& path, size_t maxBytes) {

There's probably a folly helper

---

src/admin/ConnectionLogsHandler.cpp line 70 at r1 (raw file):

}

folly::coro::Task<std::unique_ptr<folly::IOBuf>> readFileTask(std::string path, size_t maxBytes) {

Is co_invoke a different way to wrap a sync task as a coro on another exec?

---

src/admin/ConnectionLogsHandler.cpp line 125 at r1 (raw file):

              .status(503, proxygen::HTTPMessage::getDefaultReason(503))
              .header("Content-Type", "application/json")
              .body(folly::IOBuf::copyBuffer("{\"error\":\"that log type is not configured\"}\n"))

Or maybe there just haven't been any connections yet?

---

src/admin/ConnectionLogsHandler.cpp line 175 at r1 (raw file):

                    if (!token.isCancellationRequested()) {
                      proxygen::ResponseBuilder(ds)
                          .status(500, proxygen::HTTPMessage::getDefaultReason(500))

I wonder if we want a generic utility in a common header like sendErrorResponse(code, body)

Also I see you are using json bodies -- do we do that elsewhere?

---

test/CMakeLists.txt line 6 at r1 (raw file):

# Increase the timeout for gtest_discover_tests to prevent flaky build failures
# when linking heavy executables under load.
set(CMAKE_TEST_DISCOVERY_TIMEOUT 60)

Why does test discovery do linking and why does it take 60s? What is the current timeout?

---

test/CMakeLists.txt line 53 at r1 (raw file):

  moqx_test_main
)
gtest_discover_tests(moqx_relay_test DISCOVERY_TIMEOUT 120)

This seems quite hefty.

[afrind]

Also, needs update and test failures

[akash-a-n]

@akash-a-n made 6 comments.
Reviewable status: all files reviewed, 6 unresolved discussions (waiting on afrind).

---

src/admin/ConnectionLogsHandler.cpp line 54 at r1 (raw file):

Previously, afrind wrote…

There's probably a folly helper

Updated it.

---

src/admin/ConnectionLogsHandler.cpp line 70 at r1 (raw file):

Previously, afrind wrote…

Is co_invoke a different way to wrap a sync task as a coro on another exec?

simplified to use co_invoke

---

src/admin/ConnectionLogsHandler.cpp line 125 at r1 (raw file):

Previously, afrind wrote…

Or maybe there just haven't been any connections yet?

'dir' here is the string, and we are checking if a dir name exists in the config is set

---

src/admin/ConnectionLogsHandler.cpp line 175 at r1 (raw file):

Previously, afrind wrote…

I wonder if we want a generic utility in a common header like sendErrorResponse(code, body)

Also I see you are using json bodies -- do we do that elsewhere?

Json is used in a few other API responses. I have updated it to use the new send_error util in the common file

---

test/CMakeLists.txt line 6 at r1 (raw file):

Previously, afrind wrote…

Why does test discovery do linking and why does it take 60s? What is the current timeout?

I don't think i wanted to commit any of this code, the tests were flaky and I let copilot fix it on my system back then. I have reverted.

---

test/CMakeLists.txt line 53 at r1 (raw file):

Previously, afrind wrote…

This seems quite hefty.

removed

[afrind]

The signals here are still red

[akash-a-n]

Fixed it

@akash-a-n made 1 comment.
Reviewable status: 2 of 8 files reviewed, 6 unresolved discussions (waiting on afrind).

[afrind]

@afrind reviewed 7 files and all commit messages, and resolved 6 discussions.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on akash-a-n).

[afrind]

Note: I rewrote the handler to stream the file in 64 kb chunks instead of reading it all into memory

@afrind made 1 comment.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on akash-a-n).