What is vmtrace ?

[박수정]

Hi!

What is vmtrace?
How can use this command?

[Alex Glennie]

The vmtrace program is used to enable tracing: this can be performed online without any need to reboot and requires need superuser privilege. The
tool may be run in two ways.



If run without parameters, the program will run in an interactive mode and will prompt for input: this is the easiest way to enable tracing and is highly
recommended. The syntax is:

# vmtrace



It can be useful to specify the required operation from the command line, for example if tracing is to be enabled from within a script. The syntax is:

# vmtrace -b -f

The flags parameter should be an OR of the following values:
1 = Tracing for Memory Corruption
2 = Tracing for Memory Leaks
4 = Tracing for Logging

Examples:
vmtrace -b 0x180 -f 0x1
enable tracing for 256 and 128 size buckets
log for memory corruption
vmtrace -b 0x100 -f 0x7
enable tracing for 256 size buckets
log for memory corruption, memory leaks and general logging.

[Maurice Peterse]

vmtrace establishes file descriptor fd as the trace file and returns the previous value of the trace file descriptor.
The trace descriptor is initially invalid. O
utput is sent
to the trace file by successful allocation events when flag VM_TRACE is on.


maurice

[Alex Glennie · Accepted solution]

Also :
How to Use vmtrace

By default the performance kernel does not do any tracing. However the kernel contains code for the online tracing of buckets which can be enabled
through the vmtrace program. The steps to be followed are:

1.Identify the bucket size from the dump. This can be done in two ways for memory corruption:
Given the data structure which is corrupted, find the size of the data structure which should give the bucket size
Use the kminfo.pl perl script. The WhichBucket function can be used for corruption. The Dbuckets function can be used for memory
leaks: if there is a bucket with a very large number of pages allocated for that size, it is probable that the corresponding bucket is
leaking.
For more information on how to use the kminfo.pl perl script, please refer to the documentation contained in the script itself.
2.Run the vmtrace program. This is an interactive tool which prompts for bucket sizes and the type of tracing required:
memory corruption
memory leaks
log all memory allocation/deallocation calls
NOTE: this tool should be used with caution: if tracing for memory corruption, please review the warning regarding memory issues before
tracing on multiple buckets.
3.After the dump is taken, analyze the logs using the vmtrace.pl perl script from Q4.

Memory and Performance Degradation

When tracing is enabled, there is some memory and performance degradation. When tracing for memory corruption, all sizes less than a page are
allocated a page and there is therefore an increased use of system memory (i.e. a 64 byte size allocation now uses a page of 4096 bytes). Hence if
the workload on a particular system uses (e.g.) the 256 byte bucket very heavily then there is considerable loss of memory. For normal to heavy
workloads 64MB should be sufficient. In certain cases, however, the overhead could be much higher: for example, if the system was running an
application which memory maps 10,000 segments of 32 pages or less, it could use 10,000 256 byte allocations which with vmtrace would result in
the allocation of 10,000 pages (~40 Mb). Since this is completely dependant on workload and there is no way to determine the potential impact, the
most practical way is to enable corruption tracing online and, if the system runs out of memory, add more physical memory and run it again.

When tracing for memory leaks or general tracing, then the default size for the log is 2MB each. The allocation size remains the same and so there is
no memory issue as in the case of tracing memory corruption. So additional memory is never needed. The performance degradation due to logging
should be negligible.



Known Problems:

In the HP-UX 10.20 Release, vmtrace can cause panics on MP systems. Patches PHKL_8376 (Series 700) or PHKL_8377 (Series 800)
should be installed.
In HP-UX 10.30 and HP-UX 11.0, tracing for leaks can cause panics on MP systems. The work around is to trace for both corruption and
leaks, rather than tracing for leaks alone. This problem has been fixed for HP-UX 11.0 with PHKL_17038 but there is no equivalent patch
for HP-UX 10.30.

NOTE: PHKL_17038 is always required when running vmtrace on HP-UX 11.00 systems.

vmtrace was developed and is maintained by the VM team in the HP-UX Kernel lab, I don't know if it is freely distributable