is lunix really limited to 32000 processes?

 

I've just bought a server with 1.5TB of RAM.  At 1-2MB per process, that would be running a million processes.  That might not make sense with only a few dozen cores, but running more than 2^15, or ~30k, processes, which, as far as I can tell is the current limit in Linux, must be useful sometimes.

According to a part of something calling itself "The Open Group Base Specifications Issue 7" [0] (I'm not sure, but this may be stuff common to both the C and POSIX standards)  :

  • blksize_t, pid_t, and ssize_t shall be signed integer types.

The relevant one is pid_t.  It doesn't place a constraint on the size.  What would break if we had a system with 32 or 64 bit signed pid_t?  (Is it common for client code to make wrong assumptions about this?)

[0] https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/sys_types.h.html

What error do we currently get if the system runs out of pids?

My basic idea is that the multi-user nature of the existing Unix design (processes, UIDs, GIDs, filesystem areas, multiple addresses and ports and network interfaces, and so on) should be fully exploited before resorting to solutions like virtualisation, containers and so on.  Modularisation-by-process should also be able to assume the process table space is limited only be genuine resource constraints, not by an arbitrary and low limit.

Comments

Post a Comment

Popular posts from this blog

the persistent idiocy of "privileged ports" on Unix

TCP and UDP have ports.  These are 16 bit; there are 65535 or so per IP address. These protocols don't care to differentiate between the ports.  Elsewhere, IANA presumes to operate a process to allocate "well-known" ports in the range 1-1023, "registered ports" in the range 1024-49151, and to reserve the remainder, 49152–65535, for "ephemeral" ports.  The caller end has to have a port, which is how replies get back within the virtual connection, and these are conventionally picked from the ephemeral range by the OS's networking stack. The whole idea of ports is ridiculous, because it allows ISPs to arse around presuming to decide which services they will "not allow".  Anything that allows IPSs to do anything other than shift opaque packets will allow ISPs to meddle and break things, and due to the Law of Meddling, if they can, they will.  I am currently working around an issue with Claro, a pretend ISP, blocking port 5060, allocated to SI
Read more

google is giving more and more 500 errors

Image
Total state capture and moral bankruptcy are not the only problems facing Google.  They are also giving an increasing rate of 500 Internal Server errors.  This is the sort that means they've messed up, but not in a way that they should have messed up.  On the other hand, they have drawn a lovely picture of a robot falling apart. Today's example is from Google Calendar under calendar.google.com.  I was trying to make an appointment, and now I'm doing this. I remember, in 2018, a friend trying to convince me that 500 Internal Server Errors no longer happened, or if they did it was my fault, and that they hadn't seen any for years, and some other things.  They assured me that they would love to see screenshots of such a rare specimen, which I bored them with for a while.  That friend happens to be a Google employee. Well, there are plenty of others posting about Google Calendar on twitter today, including screenshots of 500 Internal Server errors, so I doubt this one is en
Read more

Guernsey Waste in incorrect bag-rejection horror May 6th, 2024

Image
The rubbish collection folk at my new place seem pickier than at other places on Guernsey. Last week, they removed some card packaging that I'd included in the card and paper (plastic) clear bag, and left it on the lane.  This week, they have attached a rejection label entitled "Polite Notice".  Just like modern software-engineered error messages (see attached tag "extremely_general_error_handling"), this does not give a specific reason why the bag was rejected, but instead lists miscellaneous possible "reasons", leaving the victim guessing, or perhaps trying to follow a process of elimination. In this case, some possible reasons, which the label does not claim to be exhaustive ("was most likely because ..."), and of which none apply, are: (i) "did not have the required payment sticker" -- no, I applied to 90 litre sticker; or  (ii) (a) "too heavy" -- no, I estimate it was 3kg, and I easily supported the bag's weight o
Read more