Uploaded image for project: 'Apache NiFi'
  1. Apache NiFi
  2. NIFI-7222

FetchSFTP appears to not advise the remote system it is done with a given resource resulting in too many open files



    • Bug
    • Status: Resolved
    • Major
    • Resolution: Fixed
    • None
    • 1.12.0, 1.11.4
    • Extensions
    • None


      Hi guys,

      We have an issue with the FetchSFTP processor and the max open file descriptors. In short, it seems that the FetchSFTP keeps the file open “forever” on our Synology NAS, so we are reaching always the default max open files limit of 1024 from our Synlogy NAS if we try to fetch 500’000 small 1MB files (so in fact it’s not possible to read the files as everything is blocked after 1024 files).

      We found no option to rise the limit of max open files on the Synology NAS (but that’s not NiFi’s fault 😉). We have also other linux machine with CentOS, but the behavior there isn’t exactly always the same. Sometimes the file descriptors get closed but sometimes as well not.

      Synology has no lsof command, but this is how I’ve checked it:

      user@nas-01:~$ sudo ls -l /proc/<SSHD SFTP process PID>/fd | wc -l


      Any comments how we can troubleshoot the issue?

      Cheers Josef

      Oh sorry, missed one of of the most important parts, we are using a 8-node cluster with nifi 1.11.3 – so perfectly up to date.

      Cheers Josef

      Hi Joe

      Ok, to our setup, we just bought a new powerful Synology NAS to use it as SFTP server mainly for NiFi to replace our current SFTP linux machine. So the NAS is empty and just configured for this single use case (read/write SFTP from NiFi). Nothing else is running there at the moment. Important limit is per SSH/user session ulimit -a 1024 open files max.:

      root@nas-01:~# ulimit -a

      core file size (blocks, -c) unlimited

      data seg size (kbytes, -d) unlimited

      scheduling priority (-e) 0

      file size (blocks, -f) unlimited

      pending signals (-i) 62025

      max locked memory (kbytes, -l) 64

      max memory size (kbytes, -m) unlimited

      open files (-n) 1024

      pipe size (512 bytes, -p) 8

      POSIX message queues (bytes, -q) 819200

      real-time priority (-r) 0

      stack size (kbytes, -s) 8192

      cpu time (seconds, -t) unlimited

      max user processes (-u) 62025

      virtual memory (kbytes, -v) unlimited

      file locks (-x) unlimited

      On NiFi side we are using an 8 node cluster, but it doesn’t matter whether I’m using the whole cluster or just one single (primary) node. It’s clearly visible that it’s related to the number of FetchSFTP processors running. So if I’m distributing the load to 8 nodes I’m seeing 8 SFTP sessions on the NAS and we can fetch 8x1024 files. I’m also seeing the file descriptors from each file (per FetchSFTP processor = PID) on the NAS which has been fetched by NiFi. In my understanding this files should be fetched and the file descriptor should be closed after the transfer, but this doesn’t seems to be the case in most of the times.

      As soon as I’m stopping the “FetchSFTP” processor, the SFTP session seems to be closed and all FDs are gone. So after stop/start I can fetch again 1024 files.

      So I tried to troubleshoot a bit further and here is what I’ve done in NiFi and on the NAS:

      A screenshot of text

      Description automatically generated

      So I’ve done a ListSFTP and got 2880 flowfiles, they will be loadbalanced to one single node (to simplify to test and only get 1 SFTP session on the NAS). In the ControlRate I’m transferring every 10 seconds 10 flowfiles to the FetchSFTP, that corelates directly with the open file descriptors on my NAS, as you can see below. Sometimes, and I don’t know when or why, the SFTP session will be closed and everything starts from scratch (not happened here) without any notice on NiFi side. As you see, the FDs are growing with +10 every 10sec and if I’m checking the path/filename of the open FDs I see that this are the one which I’ve fetched.

      root@nas-01:~# ps aux | grep sftp

      root 1740 0.5 0.0 240848 8584 ? Ss 15:01 0:00 sshd: ldr@internal-sftp

      root 1753 0.0 0.0 23144 2360 pts/2 S+ 15:01 0:00 grep --color=auto sftp

      root 15520 0.0 0.0 241088 9252 ? Ss 13:38 0:02 sshd: ldr@internal-sftp


      root@nas-01:~# ls -l /proc/1740/fd | wc -l


      root@nas-01:~# ls -l /proc/1740/fd | wc -l


      root@nas-01:~# ls -l /proc/1740/fd | wc -l


      root@nas-01:~# ls -l /proc/1740/fd | wc -l


      root@nas-01:~# ls -l /proc/1740/fd | wc -l


      root@p-li-nas-01:~# ls -l /proc/1740/fd | head

      total 0

      lr-x------ 1 root root 64 Mar 4 15:01 0 -> pipe:[1086218]

      l-wx------ 1 root root 64 Mar 4 15:01 1 -> pipe:[1086219]

      lr-x------+ 1 root root 64 Mar 4 15:01 10 -> /volume1/test/2019-08-31/detail-20190831-0104-92.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 100 -> /volume1/test/2019-08-31/detail-20190831-0052-91.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 101 -> /volume1/test/2019-08-31/detail-20190831-0340-92.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 102 -> /volume1/test/2019-08-31/detail-20190831-0246-91.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 103 -> /volume1/test/2019-08-31/detail-20190831-0104-91.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 104 -> /volume1/test/2019-08-31/detail-20190831-0150-91.log.gz

      lr-x------+ 1 root root 64 Mar 4 15:03 105 -> /volume1/test/2019-08-31/detail-20190831-0013-91.log.gz

      So to sum up, one FetchSFTP generates one SFTP Session on the NAS. The SFTP Session holds FDs which ,most of the time, doesn’t get closed. Reproduceable with the template above with FetchSFTP to a CentOS machine or a Synology NAS. Main question is now, why were the FDs not closed or when should the SFTP session gets closed.


      Just checked the open sockets on the NiFi machine where FetchSFTP is running, of course there is just one SSH session if I’m using just one single processor… So the SFTP transfer is hidden in the SSH session.

      Open TCP sessions on NiFi
      [user@ nifi-05 ~]$ netstat -vatn | grep x.y.z.232

      tcp 0 0 x.y.z.144:33628 x.y.z.232:22 ESTABLISHED

      Any comments are welcome. Still unclear where the open FDs on the NAS (SFTP server) are coming from or how it should work from NiFi perspective.

      Cheers Josef


        Issue Links



              joewitt Joe Witt
              joewitt Joe Witt
              2 Vote for this issue
              10 Start watching this issue



                Time Tracking

                  Original Estimate - Not Specified
                  Not Specified
                  Remaining Estimate - 0h
                  Time Spent - 0.5h