Details
-
Type:
Bug
-
Status: Done (View workflow)
-
Priority:
Highest
-
Resolution: Done
-
Affects Version/s: None
-
Fix Version/s: Hyrax 1.15.1
-
Labels:None
-
Epic Link:
Description
Based on the error logs, it could be here, since this is the only call to curl_easy_strerror() that is not prefixed with some identifying string.
CURLcode res = curl_easy_setopt(handle->d_handle, CURLOPT_URL, chunk->get_data_url().c_str());
if (res != CURLE_OK) throw BESInternalError(string(curl_easy_strerror(res)), __FILE__, __LINE__);
inside
dmrpp_easy_handle *
CurlHandlePool::get_easy_handle(Chunk *chunk)
{
}
But that is not the case. It's actually
2018-10-04T00:37:52GMT|&|1863|&|ERROR: BES Internal Error: Too many open files (BES error code: 1). 2018-10-04T00:37:52GMT|&|1863|&| at: BESFileLockingCache.cc:493
(I changed how BESError objects are logged when Verbose logging is turned on.)
This is error comes from:
static bool getSharedLock(const string &file_name, int &ref_fd) { BESDEBUG("cache2", "getSharedLock(): Acquiring cache read lock for " << file_name <<endl); int fd; if ((fd = open(file_name.c_str(), O_RDONLY)) < 0) { switch (errno) { case ENOENT: return false; default: throw BESInternalError(get_errno(), __FILE__, __LINE__); } } ... }
I will change the BESDEBUG lines so that we can track lock/unlock operations and re-run the tests.
